University of Southampton Research Repository ePrints Soton
Copyright © and Moral Rights for this thesis are retained by the author and/or other copyright owners. A copy can be downloaded for personal non-commercial research or study, without prior permission or charge. This thesis cannot be reproduced or quoted extensively from without first obtaining permission in writing from the copyright holder/s. The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the copyright holders.
When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given e.g.
AUTHOR (year of submission) "Full thesis title", University of Southampton, name of the University School or Department, PhD Thesis, pagination
http://eprints.soton.ac.uk
UNIVERSITY OF SOUTHAMPTON
FACULTY OF NATURAL AND ENVIRONMENTAL SCIENCES
Chemistry
DEVELOPMENT OF RESONANT INELASTIC X-RAY SCATTERING SPECTROSCOPY FOR 4d AND 5d TRANSITION METAL CATALYSTS
Rowena Thomas
Thesis for Degree of Doctor of Philosophy
APRIL 2013 UNIVERSITY OF SOUTHAMPTON
ABSTRACT
FACULTY OF NATURAL AND ENVIRONMENTAL SCIENCES
Chemistry
Doctor of Philosophy
DEVELOPMENT OF RESONANT INELASTIC X-RAY SCATTERING SPECTROSCOPY FOR 4d AND 5d TRANSITION METAL COMPLEXES
By Rowena Thomas
This research focuses on the development of Resonant Inelastic X-ray Scattering spectroscopy (RIXS) as a tool in homogeneous catalysis for 4d and 5d transition metals. In the RIXS data 2D plots of x-ray emission spectra as a function of absorption were obtained, showing the relationship between the two techniques as well as probing both the unfilled and filled DOS. They also provided L edge spectra with greatly reduced lifetime broadening. Previous studies have shown the L-edge x-ray absorption near edge structure (XANES) to be sensitive to the oxidation state and geometry, but the origins of spectral features are not always well understood. The aim of this work was to use RIXS to gain a more detailed understanding of these features and the electronic and geometric information that can be deduced from the spectra.
Molybdenum core to core RIXS data for a series of reference compounds was successfully analysed with the help of simulations using an extension of FEFF9. This showed the potential for the use of RIXS in determining more accurate oxidation states and deriving information about the geometry. Crystal field splitting parameters could be extracted directly from the dd band splitting observed.
Novel high energy resolution XANES and valence band RIXS data has been obtained for a series of tungsten and rhenium reference compounds, and the spectra have been simulated using an extension of the FEFF9 multiple scattering code. Clear trends in the incident energy and energy transfer position can be seen as a function of oxidation state and ligand type. This information was then applied to interpret the VB RIXS obtained on a homogeneous tungsten dimerisation catalyst, and used to provide insights into the oxidation state and ligand type of the catalytic intermediates before, during and after catalysis.
ACKNOWLEDGEMENTS The last four and a half years have been full of highs and lows; I couldn’t have got through this time without the help of my friends and colleagues.
Firstly I would like to thank my supervisor Moniek Tromp for all her support, advice, opportunities and encouragement over the course of my PhD, it has been a pleasure to work with you. I would also like to thank Andreas Danopoulos for all his help and supervision during the first two years of my PhD, thank you for making lab work so fun and for sharing your passion for Chemistry. I would also like to thank John Evans for all of his helpful advice and support.
I am grateful for all the help and advice from different collaborators and beamline staff during the course of my PhD. I would like to thank Josh Kas for all his invaluable assistance with the FEFF calculations and discussions on the related theory. I would also like to thank Pieter Glatzel, John Rehr and Frank de Groot for their useful input on RIXS theory. For all their help with the HERFD and RIXS experiment I would like to thank the staff at ID26, ESRF, in particular Pieter Glatzel, Janine Grattage and Kristina Kvashnina. Thanks to Evgueni Kleimenov for assistance at the SLS.
I would like to thank all the members of the Tromp/Evans group during my time at Southampton. In particular I would like to thank Sarah Hobbs for being an incredibly helpful colleague and a wonderful friend. It made all the difference to have someone to share the pain and make beamtimes fun! To Peter Wells I genuinely appreciate all the advice, encouragement and inappropriate humour over the years. Thank you to Stuart Bartlett for all the help on beamtimes and for always making the office more fun when you were around! In addition thank you to Khaled Mohammed and Michal Perdjon-Abel for all your kind words and encouragement.
Elsewhere in the Chemistry department there are a number of people I am grateful to, but especially the following. Thank you to Paolo Farina for being generally wonderful and providing much tea and hugs in stressful times. Christianne Wicking thank you for being an amazing friend and for always being right. Sophie Benjamin thank you for always brightening up my day! Thanks to Susana Conde-Guandano, Alan Henderson, David Bolien and Cyril Henry for making the lab so much fun (and so efficient!) in my first two years.
Finally I would like to thank my parents and my favourite sister for all their love and support, for always believing in me, and for knowing not to ask when I would finish this thesis.
Glossary of Terms Techniques
XAS X-ray Absorption Spectroscopy
XES X-ray Emission Spectroscopy
XAFS X-ray Absorption Fine Structure
EXAFS Extended X-ray Absorption Fine Structure
XANES X-ray Absorption Near Edge Structure
RIXS Resonant Inelastic X-ray Scattering
VB RIXS Valence Band Resonant Inelastic X-ray Scattering
DFT Density Functional Theory
LDOS Local Density of States
DOS Density of States
MS Multiple Scattering
MO Molecular Orbital
NMR Nuclear Magnetic Resonance Spectroscopy
IR Infrared
UV/vis UV- visible
General
ν Vibrational frequency
λ Wavelength
Hz Hertz
δ Chemical shift ppm Parts per milliion s Singlet d Soublet
t Triplet m Multiplet fac Facial mer Meridional
Me Methyl
Et Ethyl
Bu Butyl
Ph Phenyl
Ar Aromatic o ortho
TABLE OF CONTENTS
Chapter 1: Introduction 1.1 Overview of project 1 1.2 Catalysis 3 1.2.1 Overview 3 1.2.2 Polymerisation of alkenes 5 1.2.3 Tungsten imido catalyst for selective dimerisation of low 7 molecular weight alkenes 1.3 XAFS techniques in catalysis 12 1.3.1 Practical developments 12 1.3.2 Theoretical considerations 14 1.4 References 16
Chapter 2: X-ray Techniques 2.1 Introduction 19 2.2 X-ray Absorption Spectroscopy 19 2.2.1 X-rays and the Photoelectric Effect 19 2.2.2 Synchrotrons and Beamlines 20 2.2.3 Detection methods 23 2.2.4 X-ray Absorption Fine Structure Spectroscopy 25 2.2.5 Extended X-ray Absorption Fine Structure (EXAFS) 28 Spectroscopy 2.3 X-ray Absorption Near Edge Structure 30 2.3.1 Introduction 30 2.3.2 Multiplet effects 30 2.3.3 Studies of K-edge XANES 31 2.3.4 L-edge XANES 32 2.3.5 Molybdenum XANES 33
2.3.6 Tungsten and rhenium XANES 35
2.3.7 Energy resolution of XANES 36
2.3.8 High Energy Resolution Fluorescence Detected XANES 36
2.4 X-ray Emission Spectroscopy 37
2.5 Resonant Inelastic X-ray Scattering 40
2.5.1 Introduction 40
2.5.2 The RIXS process 41
2.5.3 Valence Band RIXS 44
2.6 Data Analysis 45
2.6.1 Summary 45
2.6.2 FEFF9 46
2.7 References 47
Chapter 3: Experimental
3.1 Introduction 53
3.2 Instrumentation and general techniques 53
3.3 Synthesis of tungsten reference compounds 54
3.3.1 WOCl 4 54
3.3.2 WCp*Me 4 54
t 3.3.3 WCl 4N(4- BuPh) 55
3.4 Synthesis of rhenium reference compounds
3.4.1 [Re(o-pda) 3][ReO 4] 56
3.4.2 Re(o-pda) 3 56
3.5 Tungsten imido catalysis 56
3.6 RIXS experiments 57
3.6.1 Experimental set-up December 2009 (ESRF) 57
3.6.2 Experimental set-up December 2010 (ESRF) 58
3.6.3 Experimental set-up July 2011 (ESRF) 58
3.6.4 Experimental set-up July 2010 (SLS) 59
3.6.5 Sample preparation – powders 59
3.6.6 Sample preparation – solutions 59
3.7 FEFF9 calculations 59
3.7.1 Overview 59
3.7.2 FEFF parameters for Mo L α RIXS calculations 60
3.7.3 FEFF parameters for W and Re L 3 valence band RIXS 62 calculations 3.8 References 67
Chapter 4: Molybdenum Reference Studies
4.1 Abstract 69
4.2 Introduction 69
4.3 Experimental 73
4.4 Results 75
4.5 Discussion 81
4.6 Conclusions 83
4.7 Notes 83
4.8 References 84
Chapter 5: Tungsten Reference Compounds
5.1 Introduction 87
5.2 Tungsten (IV) oxide 88
5.3 Tungsten (VI) oxide 93
5.4 Lithium tungstate 97
5.5 Tungsten (VI) oxotetrachloride 99
5.6 Tungsten hexacarbonyl 103
5.7 Bis(tert-butylamido) bis(tert-butylimido) tungsten 107
5.8 Pentamethylcyclopentadienyl tetramethyl tungsten 110
5.9 Dimethoxyethane dichlorodioxo tungsten 113
5.10 WCp*Cl 4 116
5.11 Tungsten hexachloride 117
5.12 Discussion and comparison of Valence Band RIXS 118
5.13 L beta 2 RIXS for tungsten reference compounds 120
5.14 Conclusions 122
5.15 References 124
Chapter 6: Rhenium Reference Compounds
6.1 Introduction 125
6.2 Rhenium (IV) oxide 125
6.3 Rhenium (VI) oxide 129
6.4 Mer - trichlorotri(methyldiphenylphosphine) rhenium 133
6.5 Trichlorooxobis(triphenylphosphine) rhenium 136
6.6 Rhenium tris(o-phenylenediamide) 139
6.7 Summary 142
6.8 Conclusions 143
6.9 References 143
Chapter 7: Tungsten Catalysis
7.1 Introduction 145
7.2 Experimental results – VB RIXS 147
7.2.1 Experimental L 3 VB RIXS spectra 147
7.2.2 Summary table of experimental L 3 VB RIXS results 150
7.3 Experimental results – HERFD reactions 152
7.4 Other experimental results 155
7.4.1 Precursor vs. in situ 155
7.4.2 Reaction with no substrate 155
7.4.3 Reaction with 6 equivalents aniline 156
7.5 Discussions and conclusions 157
7.6 References 158
Chapter 8: Conclusions and recommendations for further work
8.1 Introduction 161
8.2 Benefits of RIXS 161
8.3 Comparison of experiment and theory 162
8.4 Catalysis 163
8.5 Recommendations for future work 164
8.6 References 165
Declaration of Authorship
I, Rowena Thomas, declare that the thesis entitled:
RESONANT INELASTIC X-RAY SCATTERING STUDIES OF 4d AND 5d TRANSITION METAL COMPLEXES and the work presented in it are my own and has been generated by me as the result of my own original research.
I confirm that:
1. This work was done wholly or mainly while in candidature for a research degree at this University;
2. Where any part of this thesis has previously been submitted for a degree or any other qualification at this University or any other institution, this has been clearly stated;
3. Where I have consulted the published work of others, this is always clearly attributed;
4. Where I have quoted from the work of others, the source is always given. With the exception of such quotations, this thesis is entirely my own work;
5. I have acknowledged all main sources of help;
6. Where the thesis is based on work done by myself jointly with others, I have made clear exactly what was done by others and what I have contributed myself;
7. None of this work has been published before submission.
Signed:…………………………………..
Date: 11/04/2013
Chapter 1 Introduction
1 Introduction 1.1 Overview of Project High energy Resonant Inelastic X-ray Scattering (RIXS) is a powerful and relatively new technique used to simultaneously probe the unoccupied and occupied electronic density of states. RIXS spectra therefore contain a wealth of electronic and geometric information, but whilst much progress has been made with the experimental technique, the theory is currently lacking in comparison.
The main aims of this project were to further develop both the experimental technique as a tool in catalysis, and the understanding of RIXS spectra in order to obtain detailed electronic characterisation of transition metal compounds. This technique has previously been used to study some 3d transition metal compounds but has not been fully explored, and there are very few studies on the 4d and 5d transition metals.
To fulfil the primary aim of this project a series of molybdenum, tungsten and rhenium compounds covering a range of oxidation states, ligands and geometries were synthesised or bought commercially where available. These were then characterised using RIXS and complementary techniques. This data was then analysed using FEFF9, a single electron multiple scattering code, to determine different orbital contributions and to assign RIXS features where possible.
In addition to these reference compounds preliminary work was undertaken on obtaining RIXS spectra of a homogeneous catalytic system in action. The long term aim of this research is to apply the developed techniques to single site catalytic systems in situ in order to gain detailed information about their changing electronic properties during catalysis. This should allow improved insights into the structure- performance and electronic-performance relationships and facilitate the design of new performance optimised catalytic systems.
4d and 5d transition metals were chosen in order to extend the RIXS technique from the 3d metals, which had been more widely studied. In addition, the L edges for 4d and 5d transition metals are accessible experimentally, allowing direct probing of the valence d-orbitals, particularly interesting for catalysis. The main advantages to using RIXS over measuring a direct L edge are that we gain 2D information and greater energy resolution.
1 Chapter 1 Introduction
In Chapter 1 the background to the project is discussed, with a brief overview of catalysis and details of the catalytic system which was examined in this project. There will also be a brief introduction to modern x-ray absorption techniques which are particularly relevant to catalysis.
In Chapter 2 the x-ray spectroscopic techniques which form the basis of this thesis will be discussed in detail, primarily x-ray absorption near edge structure (XANES) spectroscopy and resonant inelastic x-ray scattering (RIXS) spectroscopy, as well as the relevant background.
In Chapter 3 the experimental work is described, the synthesis of any compounds that are described in this thesis, details of the synchrotron based experiments and details of the FEFF9 calculations that were used to simulate and help analyse the XANES and RIXS spectra.
Chapter 4 discusses molybdenum L α RIXS spectra for a series of reference compounds. FEFF9 was used to produce simulations of the 2D RIXS contour plots. Additionally local density of states plots were calculated to show the relative occupied and unoccupied orbital contributions, in order to help fully assign the features in the RIXS plots.
Chapters 5 discusses valence band RIXS for a series of tungsten reference compounds, along with the results of the RIXS and local density of states calculations performed using FEFF9.
Chapters 6 presents valence band RIXS for a series of rhenium reference compounds, along with the results of the FEFF9 calculations performed for these materials.
Chapter 7 discusses the valence band RIXS spectra and the high energy resolution fluorescence detected (HERFD) XANES spectra that were collected for a catalytic system. The catalyst studied was a homogeneous tungsten imido system formed in- situ to catalyse the selective oligomerisation of alkenes.
Chapter 8 details the conclusions that have been made based on the work described in this thesis and future recommendations.
2 Chapter 1 Introduction
1.2 Catalysis 1.2.1 Overview A catalyst is a material which, in relatively small amounts, speeds up the rate of a chemical reaction. Unlike the reactants the catalyst is not consumed by the reaction. The catalyst may be changed during the reaction but by the end it will be in its original composition and able to repeat the process 1. Catalysts are of huge economic value, they are used in the production of many chemicals and petroleum products. They are also increasingly contributing to a cleaner environment, both through reducing by- products in industrial reactions and reducing pollutants in systems such as the catalytic converters in vehicle exhaust systems 1.
Catalysts increase the rate of a reaction by reducing the activation energy needed, i.e. the minimum energy required to start a chemical reaction. The catalyst does this by offering a new pathway with a lower Gibb’s energy of activation. However the Gibb’s energy of the overall reaction is not changed, so a reaction will not be made more or less thermodynamically favourable by a catalyst 1. It is also possible for catalysts to act selectively, which will make one product more favourable whilst minimising side products 1,2. This ability to only increase the rate of certain reactions is extremely valuable, enabling a chemical process to progress more efficiently and often with less waste, making catalysts valuable to industry 1. The efficiency of a catalyst is also a key consideration, which can be measured by the turnover frequency (or turnover number). A catalyst with a high turnover frequency will be highly active, even in low concentrations 1.
We can divide catalysts into two main categories; heterogeneous and homogeneous. A heterogeneous catalyst is one which is in a different phase from the reactants, i.e. a solid catalyst and gaseous reactants, whereas a homogeneous catalyst is in the same phase as the reactants. One downside to homogeneous catalysis is that it is necessary to separate the catalyst at the end of the process, which can be a difficult and time- consuming process. This is one reason why heterogeneous catalysts are more widely used in industry, along with their higher stability at high temperatures. When considering the spectroscopic analysis of heterogeneous catalysts surface effects need to be accounted for, whereas in homogeneous catalysis we typically consider a uniform bulk system, which can simplify the analysis 1. Homogeneous catalysts will also have a well defined starting point when considering structural analysis. Homogeneous catalysts are also often more active and more selective 1.
3 Chapter 1 Introduction
During the catalytic process, for both homogeneous and heterogeneous catalysts, at least one of the reactants will associate with the catalyst at its active sites. There will then be an interaction between the catalyst and reactant molecules which will make them more reactive. This could be reactions at the surface or weakening of bonds in the reactant molecules. Then the reaction will occur; at least one of the reactant molecules will be attached to the catalyst, the other may also be attached or be moving through the liquid or gas until it hits the molecule bound to the catalyst. The product molecule(s) will dissociate from the catalyst, leaving the active site vacant for more molecules to react 1. Other molecules can bind to or interact with a catalyst other than the reactant molecules. Inhibitors are substances which will slow down the effect of a catalyst, whereas promoters are substances that increase their effect. Catalytic poisons are substances which deactivate the catalyst 1.
Catalysts can be difficult to study spectroscopically as most intermediate species are highly reactive and are often present in very low concentrations. Therefore, mechanisms cannot often be determined with complete certainty and new developments in characterisation can add valuable insights 1. Techniques such as UV- visible and infrared (IR) spectroscopy are commonly used in the characterisation of catalysts in-situ . In UV-vis spectroscopy photons in the ultraviolet and visible regions of the electromagnetic spectrum are absorbed by compounds, resulting in electronic transitions within the molecules. It is a commonly used technique in the quantitative analysis of transition metals in solution and organic compounds with high levels of conjugation.
IR spectroscopy, in conjunction with complementary techniques, can be very useful for both qualitative and quantitative analysis of molecules. IR spectroscopy probes the vibrational modes of molecules and the spectra will contain characteristic bands which correspond to different functional groups. Specific functional groups will absorb infrared radiation at specific frequencies so are easily identifiable. However, these characteristic functional group frequencies will be affected by their environment. For example, changes in geometry due to steric effects and the presence of neighbouring groups of differing electronegativity can affect the bond angles 3 and, thereby, affect the position of the bands. The intensity of the bands can be attributed to a range of different factors, including how often the functional group is present in the molecule, the solvent and the neighbouring atoms. Therefore, although useful for known reference compounds, for
4 Chapter 1 Introduction unknown compounds complementary information is vital due to the variety of different factors which govern the IR spectra.
In this study we will be using x-ray absorption and emission techniques to probe the catalyst. We will use these techniques to look at a homogeneous catalytic system, that of a tungsten imido complex, formed in-situ , which catalyses the selective dimerisation of α-olefins.
1.2.2 Polymerisation of alkenes The polymerisation of short chain olefins is an important process in industry; which forms higher molecular weight alkenes that are of great use as industrial intermediates 4. The polymerisation of alkenes can produce useful and widely used polyalkenes 1. For example polyethene, an addition polymer formed from ethene molecules, is one of the most common plastics and is commonly used for packaging. There are different varieties of polyethene. Low density polyethylene (LDPE) consists of chains with a large amount of branching, which results in a significant amount of amorphous regions. In these less well ordered regions Van der Waals forces are less effective resulting in a low strength but flexible low density polymer. It is commonly used for carrier bags. High density polyethylene (HDPE) consists of chains which have very little branching which results in better packing and therefore more effective Van der Waals interactions. Therefore HDPE is stronger than LDPE, with a higher density and a higher melting point. It has many uses including plastic pipes and plastic milk bottles. Unlike LDPE, which is formed in a high temperature and high pressure radical process, HDPE and also linear low density polyethylene (LLDPE) are manufactured using homogeneous or heterogeneous catalysts at relatively low temperatures (80- 180°C) and pressures (less than 50 bar) 5. Typically Ziegler-Natta or Phillips type catalysts are used (see below).
Polychloroethene, more commonly known as polyvinyl chloride (PVC), consists of amorphous chains which would usually result in a relatively flexible and low strength polymer. However the polar carbon to chlorine bonds result in dipole-dipole interactions which make PVC hard and rigid, to increase flexibility plasticisers can be added. The uses for PVC include electrical cable insulation, plastic windows and clothing. Polytetrafluoroethene (PTFE) is a polymer with a high melting point and is very chemically resistant; it is used as a non-stick coating for pans and related items.
5 Chapter 1 Introduction
Ziegler discovered some of the first examples of catalysts for such polymerisation reactions in the 1950s with his series of organoaluminium compounds coupled with nickel reagents, which were able to selectively polymerise alkenes at low temperatures and pressures 4,6. Since this time many transition metal catalysts have been developed for the polymerisation of ethylene and propylene. The most likely proposed mechanism for the Ziegler-Natta catalyst is the Cossee-Arlman mechanism, shown in Figure 1.1.
To produce the catalyst TiCl 4 and AlEt 3 are reacted to form a fine powder of polymeric
TiCl 3. The triethylaluminium will alkylate a surface Ti atom and an ethane molecule will co-ordinate to a neighbouring vacant site, shown by the grey circle in Figure 1.11. The co-ordinated alkene then undergoes a migratory insertion reaction, opening up another vacancy so this process can be repeated and the chain can grow. The polymer is released from the Ti by β-hydride elimination, terminating the chain 1.
Cl Cl Cl Cl Chain Ti Ti growth Cl Cl Cl
Cl Cl Cl Cl Ti Ti Cl Cl
Migratory Cl Cl insertion Ti Cl
Figure 1.1 - Cossee type mechanism for Ziegler-Natta catalysts in
polymerisation of ethene 1
Another industrially important catalyst is the Phillips catalyst, which is used in the production of more than one-third of the PE sold globally 5,7. This type of catalyst, first patented in 1958 8, is a supported chromium oxide catalyst, usually prepared by impregnating wide pore silica with a chromium oxide7. Polymerisation by Phillips type
6 Chapter 1 Introduction catalysts is successful for a diverse range of polymers; it is able to make over 50 different types of HDPE and LLDPE 5. No activator is needed for catalysis using Phillips type catalysts, unlike the Ziegler process, which simplifies the industrial process 5. There is still much controversy over the structure of the active site and the initiation process, in particular.
The third main type of polymerisation catalysts are single-site homogeneous catalysts, or supported homogeneous catalysts, for example metallocenes. These are generally bis-cyclopentadienyl derivatives of group 4 metals (Zr, Hf, Ti)7,9. These are very active catalysts, with turn over frequencies (TOF) of up to 240 000 per hour 10 for ethylene trimerisation and they produce up to about 90% of the desired product, in this case 1- hexene 10 .
1.2.3 Tungsten imido catalyst for selective dimerisation of low molecular weight alkenes The use of homogeneous catalysts in the dimerisation of α-olefins is an attractive route to the formation of high molecular weight olefins. Homogeneous catalysts are usually much more selective than heterogeneous catalysts, and often have very high activities due to the accessibility of catalyst molecules in solution 1. The majority of the existing catalysts produce branched molecules; however, certain late transition metal complexes (for example nickel, cobalt 11 ) have been successful in linear oligomerisation 11 . There are far fewer examples for group 6 metals, with the chromium catalysed trimerisation of ethylene 12 the most successful case previously. However recent developments have shown tungsten imido complexes to be highly selective and moderately active 16,17,20 . More specifically tungsten mono(imido) complexes, when treated with Lewis acids of the form AlCl nRm, have been reported as efficient catalysts for the dimerisation of α-olefins 13 . A similar example has also been reported for tungsten bis(imido) compounds 14 . Whereas a large excess of activator is usual for oligomerisation catalysts here a specific amount of ~15 molar equivalents is needed, indicating a particular function in the catalysis in addition to simple activation. An overview of this reaction is shown in Figure 1.2.
7 Chapter 1 Introduction
Figure 1.2 - Dimerisation of α-olefins with a tungsten based catalyst formed in-situ 17
This catalyst was first mentioned by Lawson, Menapace et al, from Goodyear, in the mid-1970s, when they observed that a WCl 6 and aniline catalyst used for olefin metathesis could also be used for catalysing the dimerisation of olefins with some adjustments 15 . By increasing the amount of the aluminium reagent used in the reaction they changed the pathway from metathesis to dimerisation 15 . In the work of Olivier and Laurent-Gérot 16 they examine the Goodyear catalyst in reactions carried out in an ionic liquid, with the aim of increasing the turn-over number using the two phase concept 16 .
The hypothesised W(VI) complex formed in-situ with the addition of aniline to WCl 6 was of the form Cl 4W=NAr, but was not isolable. So they synthesised complexes of this form and reacted them with ethylalumnium dichloride (EADC), which yielded catalytic results comparable to those obtained by the in-situ method. More recently this catalyst system has been the subject of a patent 13 and subsequent papers investigating the mechanism and selectivity 14, 17, 20 . These studies differed from the previous work on this type of catalyst because they used a base to remove the HCl produced in the catalyst formation, rather than the sparging or excess aluminium reagent used previously 17 . All these studies indicated that a very specific number of equivalents of the aluminium activator were needed, which points towards a very specific role in the catalytic cycle as they are usually present in a large excess 18 . The catalyst is formed in-situ from tungsten hexachloride, aniline, triethylamine and ethylaluminium dichloride, although variations on the aniline and alkyl aluminium have been probed in detail, with big differences seen in the activity and selectivity of the catalyst formed 17 . Density Functional Theory (DFT) studies have also been performed in order to probe the energetically favourable mechanism and to try and understand the role of the aluminium activator 14, 20 .
8 Chapter 1 Introduction
There are two different mechanistic pathways possible for these dimerisation reactions, one being the migratory insertion of olefins into metal-hydrogen or metal-alkyl bonds followed by β-hydride elimination, typical for the late transition metal systems. The more likely method, given the high selectivity and the substitution pattern of the dimerised products, is a metallacycle mechanism. This is also the mechanism identified for chromium catalysed ethylene polymerisation 19 . This metallacycle hypothesis has been examined in a theoretical study 20 using density functional theory (DFT) which presents the likely mechanism as metallacyclic, however with no experimental data this could not be proven at the time.
The proposed catalytic cycle described by Tobisch 20 is presented in Figure 1.3 and was the focus of the DFT study in this paper. One of the main aims in Tobisch’s study was to identify the role of the Lewis acid, an alkyl aluminium dihalide, in the catalytic cycle. The catalytic cycle was first analysed in the case that the Lewis acid does not participate in the cycle, but this was found to be energetically unfavourable in some steps, particularly the degradation of 3 in Figure 1.3. Also dimerisation is significantly less likely due to the increased favourability of growth of 3 compared to degradation. Tobisch then looked at the complexation of the Lewis acid onto 3 (see Figure 1.3), looking at different sites in the immediate vicinity of the tungsten centre 20 which were of varying stabilities. From these studies it is suggested that the Lewis acid has an important role in regulating the selectivity of dimerisation.
In the 2010 study 17 by Hanton et al they carried out the catalysis using systematic variations in each of the main components in order to determine which functional groups played a vital role. They determined that WCl 6 alone does not dimerise olefins; therefore the active catalyst is likely to be a tungsten-amido or –imido species, based on the observance of HCl being produced. It was also found that an elevated temperature (ideally around 60°C) was needed for the catalyst formation process, at 20°C there does not appear to be complete formation of the pre-catalyst and only comparatively low conversion was seen. High temperatures, at 100°C and 132°C, for both the formation and catalytic temperatures, were also not optimal, giving low selectivity and low productivity.
9 Chapter 1 Introduction
Figure 1.3 - Proposed catalytic cycle for α-olefin dimerisation by mono(imido) tungsten compounds, with ethylene as substrate. Reprinted with permission from Tobisch, S., Organometallics 2007, 26 , 6529-6532. Copyright 2007 American Chemical Society.
The choice of aniline was also probed 17 , with the expected result (based on a previous report 16 ) being an increased rate with an electron withdrawing group on the aniline ring, and a decreased rate with the substitution of an electron donating group on the ring. The addition of a para-methyl group supported this trend, with a significant drop in the rate, however, the presence of a p-fluorine group had no effect on selectivity or activity of the catalyst. Therefore it does not appear that a more electrophilic tungsten atom influences the rate determining step. The analogous p-chlorine system gave poor results, with less activity than the p-methyl aniline. This was hypothesised to be due to a reaction with the ethyl aluminium dichloride species, yielding a p-ethyl group. From these results a system of WCl 6, PhNH 2 and Et 3N was chosen for further study due to good activity and high selectivity towards mono- and di-methyl branched dimers 17 . Polymers were never observed, with trimers and tetramers the only heavy products formed.
10 Chapter 1 Introduction
A number of aluminium activators were studied with this system, as this has been shown to have a significant effect on other oligomerisation catalysts 16,17,21,22 . A number of alternatives were tested and it was found that both an alkyl and a halide group are essential at the aluminium centre. Without the presence of both of these functional groups dimerisation did not occur. It was found that EADC gave superior results in terms of activity and selectivity to the dimer fraction. A comparison was made between the use of methyl aluminium dichloride (MADC) and EADC, to see if a β-hydrogen was necessary on the alkyl group, and whilst catalysis still occurred the activity was decreased significantly. However it did still occur, indicating that a β-hydrogen elimination from a tungsten alkyl group formed in-situ is not involved in the catalytic pathway. The type of halide bonded to the aluminium was also investigated, when switching from EADC to ethyl aluminium dibromide (EADB) the catalyst was much less active and much less selective, suggesting the halide has an important role. The solvent was also investigated and chlorobenzene was found to give superior results, both in terms of activity and selectivity.
Mass balance experiments showed that no polymer products were formed, and there was high selectivity to linear dimers: 95.3% which is higher than comparable catalysts 17 . Some heavier products were formed, but these were primarily trimers and tetramers. Hanton et al 17 considered the possibility of a metallacyclic versus a Cossee- type mechanism based upon the high selectivity of the catalyst. There have been examples of a metallacyclic mechanism for highly selective alkene oligomerisation chromium catalysts 23 so it would not be unprecedented. The detailed computational studies by Tobisch 14,20 indicated that a metallacyclic mechanism was energetically feasible for tungsten-imido based catalysts. No similar studies were carried out for Cossee-type pathways in these systems. By comparing the oligomerisation products to the expected isomers for both mechanisms the results showed that the catalyst appeared to be proceeding by a Cossee-type mechanism. To confirm these results with a second method an isotopomer analysis 23 was carried out using a mixture of
C2H4/C 2D4 in order to identify the mechanism based on the products. The results showed full isotopic scrambling of all the C 4 and C 6 molecules which indicated that the reaction progressed via the Cossee-type mechanism 17,24 , contradicting the metallacyclic theory proposed in the work of Tobisch 14,20 .
11 Chapter 1 Introduction
1.3 XAFS techniques in catalysis 1.3.1 Practical developments Spectroscopy covers a range of techniques designed to probe the interaction between light and matter 25 , in which the absorption and emission of the photon energy by matter is studied, in particular the dependence of these interactions on the wavelength of the light. Different energies, in different regions of the electromagnetic spectrum, will interact with atoms and molecules in different ways, yielding a range of different information; including structural parameters and electronic information. Spectroscopy has been used across a wide variety of fields in science, but we are primarily concerned here with the study of transition metal complexes of importance in catalysis. In order to improve the performance of existing catalysts, and to better design future catalysts, it is important to understand their mechanisms, in particular the structure of the active species and how this changes throughout the reaction. In metal catalysts the reactivity is governed by the metal d-orbitals and the ligand valence orbitals 26 . Therefore, it is important to understand the electronic properties of the activated species and any charge re-distribution that may be taking place 26 .
A useful method for studying the electronic and geometric properties of transition metals is X-ray Absorption Near Edge Structure (XANES) spectroscopy, which will be discussed in Chapter 2. Electronic information about the active species during a catalytic reaction would be extremely valuable in predicting catalytic behaviour, if the data from such techniques can be reliably interpreted in a quantitative manner. In recent years there have been many important experimental developments in the field of XAFS which has offered many new opportunities to study reactions in-situ. Third generation synchrotrons, offering high photon fluxes, increasingly smaller beam sizes and faster data acquisition have allowed the development of many techniques previously not possible, some of which are discussed below.
Due to the wealth of element-specific information, structural, geometric and electronic, contained within x-ray absorption spectra this technique is extremely valuable in the characterisation of materials. Increasingly it has been used to study reaction pathways and intermediates in-situ . The high penetration depth of hard x-rays means that a wide range of conditions are possible, including high temperatures and pressures. In-situ XAFS experiments can be combined with complementary techniques in order to gain more information about the system, which is very useful for studying catalytic reaction. One such method is the combination of Diffuse Reflectance Infrared Fourier
12 Chapter 1 Introduction
Transform Spectroscopy (DRIFTS), EXAFS and Mass Spectrometry (MS) to study heterogeneous catalytic reactions in-situ 27 . This ensures that the results from each technique are taken from the same experiment, therefore with equivalent conditions, this also minimises valuable time needed at synchrotrons. In this particular set-up the MS measures the components of the gas phase whilst the EXAFS and the IR probe the structure and bonding of the solid phase catalysts, giving complementary information which can be combined to give a fuller picture of the catalyst’s behaviour. It has been used to study, for example, rhodium nanoparticles on alumina supports 28 . The interactions with a variety of small gas phase molecules were studied and the surface species formed by the exposure of the nanoparticles to the absorbing molecules were probed spectroscopically as a function of both temperature and time. These combined studies show the strong effect of the absorbing species on the nanoparticles and their reactivity 28 .
Modern third generation synchrotrons, providing high photon flux, have paved the way for techniques such as quick-EXAFS (QEXAFS). In a traditional EXAFS experiment the monochromator consists of two fixed, parallel crystals which move stepwise through different angles and therefore energy points. In a QEXAFS experiment the monochromator is continuously oscillating 29 . Without the time delays needed to move the traditional motors for each point QEXAFS spectra can usually be obtained in under one minute, sometimes even on a 1s timescale 30 . This is ideal for studying catalysts in- situ because with traditional scanning EXAFS a scan would typically take at least 15 minutes, which may miss many changes in a reaction 30 . For example to follow the dynamic structural changes in Cu/ZnO catalysts Grunwaldt et al 31 used QEXAFS in combination with x-ray diffraction (XRD) and MS. QEXAFS at the Cu K-edge was used in order to determine the local environment around the copper atoms during the reaction and during the heating steps.
Another technique well suited to probing catalysts in-situ is Energy Dispersive EXAFS (EDE). Unlike scanning EXAFS experiments which use a pair of parallel crystals for the monochromator, EDE uses a bent crystal polychromator to disperse a range of x-ray energies onto the sample 30 and spectra are measured using a position sensitive detector. A spectrum can be measured on the scale of 10s to 100s of milliseconds, however often multiple scans are needed to improve the signal to noise ratios 30 . Even when measuring an average of 10s or 100s of scans it is possible to gather enough data for one spectrum in less than one second 30 . With such short scan times it is possible to capture spectra of short lived reaction intermediates. By using techniques
13 Chapter 1 Introduction such a QEXAFS or EDE to collect fast but high quality XANES spectra we can monitor changes in the electron configuration and oxidation state, and gain insights into the reaction dynamics.
Another method of probing reactions in-situ is ultrafast XAFS with a pump/probe technique, using short laser and x-ray pulses. This is a powerful technique which can be used to gather structural and electronic information about molecules with short-lived excited states 32 . The first short laser pulse photoexcites the molecule, the following x- ray pulse probes the molecule spectroscopically 32. Although this has been used primarily for photochemical reactions it shows promise for use in catalysis, to detect and characterise short lived reaction intermediates. With increasingly precise synchronisation between the laser and x-ray beam possible; more and more short-lived intermediates are detectable. In early work using a pump-probe XAFS technique Chen et al 33 determined the transient molecular structure of a photoexcited solution of
NiTPP-L2 (Nickeltetraphenylporphyrin, where L is piperidine) on a 14 nanosecond time scale. The XAFS results confirm the intermediate structure is square planar NiTPP with a lifetime of around 28 nanoseconds, following removal of the 2 axial ligands.
X-ray emission spectroscopy (XES) is of interest in catalyst because it can clearly distinguish between ligands, such as carbon, nitrogen and oxygen 34 , and in-situ dynamic measurements are possible with transitions in the hard x-ray region. Soft x-ray emission spectroscopy has been successfully used to probe the orbitals involved in bonding between adsorbates and heterogeneous transition metal catalysts 34 . For example XES studies showed that CO adsorbed on Ni (100) surfaces on on-top sites preferentially, with the carbon bonded to the surface 34 .
1.3.2 Theoretical considerations Whilst the analysis of extended x-ray absorption fine structure (EXAFS) spectroscopy is well established, the qualitative analysis of x-ray absorption near edge structure (XANES) spectroscopy is still not fully understood. However many different approaches are currently being developed in order to uncover the wealth of electronic information contained within this region. At present the theoretical techniques cannot keep up with the many experimental developments, and there is not yet agreement on the best approach, although each has contributed to an increased understanding of the edge region.
14 Chapter 1 Introduction
There are three main, broad areas of theoretical approach to analysing XANES spectra; density functional theory (DFT) 35 , multiple scattering (MS) and molecular orbital 36 based methods. DFT approaches probe the occupied density of states, whereas in XANES we are probing the unoccupied density of states. Multiple scattering is one of the most popular methods, also a popular method for analysing EXAFS data, and will be discussed in more detail in chapter 2, particularly in reference to FEFF which was used to simulate the data obtained in this project. Unlike EXAFS, where quantitative results can generally be accurately extracted, more work on the theoretical understanding of the XANES region is needed in order to extract precise chemical information. There are a number of factors which complicate the analysis of XANES spectra. For example both the lifetime broadening and the experimental broadening affect the spectra, and can blur the fine structure around the edge reason making it difficult to analyse. In addition a variety of different effects contributes to the pre-edge and edge regions, such as band effects and multiplet effects, and there is no comprehensive theory that encompasses all the effects. There are complications due to the relatively low kinetic energy of the photoelectron in the XANES region, which results in large scattering amplitudes 37 . In addition there is the problem of accounting for the core holes theoretically, in a RIXS experiment there are multiple holes at different stages of the experiment, and in different orbitals.
15 Chapter 1 Introduction
1.4 References
1 Atkins, P.; Overton, T.; Rourke, J.; Weller, M.; Armstrong, F., Chapter 25: Catalysis, Shriver and Atkins Inorganic Chemistry Fourth Edition, Oxford University Press: Oxford, 2006, 680-684.
2 Du, X.-D., Yu, X.-D., J. Mol. Catal. A. Chem., 1997, 126,109.
3 Socrates, G., Infrared and Raman Characteristic Group Frequencies. John Wiley
& Sons, LTD: London, 2001.
4 Pillai, S. M.; Ravindranathan, M.; Sivaram, S., Chem. Rev., 1986, 86 , 353.
5 Groppo, E.; Lamberti, C.; Bordiga, S.; Spoto, G.; Zecchina, A., Chem. Rev. 2005, 105 , 115.
6 Ziegler, K.; Holzkamp, E.; Breil, H.; Martin, H., Angew. Chem., 1955, 67 , 541.
7 Weckhuysen, B. M.; Schoonheydt, R. A., Catal. Today 1999, 51 , 215.
8 Hogan, J. P.; Banks, R. L. U.S. Patent 2,825,721, 1958
9 Brintzinger, H. H.; Fischer, D.; Mulhaupt, R.; Rieger, B.; Waymouth, R. M., Angew. Chem.-Int. Edit. Engl. 1995, 34 , 1143.
10 McGuinness, D. S., Chem. Rev., 2010, 111 , 2321.
11 (a) Keim, W.; Hoffmann, B.; Lodewick, R.; Peuckert, M.; Schmitt, G.; Fleischhauer, J.; Meier, U., J. Mol. Catal., 1979, 6, 79. (b) Small, B. L.; Marcucci, A. J., Organometallics 2001, 20 , 5738. (c) Small, B. L., Organometallics 2003, 22 , 3178.
12 Carter, A., Cohen, S. A., Cooley, N. A., Murphy, A., Scutt, J., Wass, D. F., Chem. Comm., 2002 , 8, 858
13 Hanton, M. J.; Tooze, R. P.; WO 2005089940 (Sasol Technology (UK) Ltd), September 29, 2005
14 Tobisch, S., Dalton Trans., 2008, 2120.
15 Menapace, H. R.; Maly, N. A.; Wang, J. L.; Wideman, L. G., J. Org. Chem. 1975, 40 , 2983.
16 Olivier, H.; Laurent-Gérot, P., J. Mol. Catal. A. Chem. 1999, 148 , 43.
17 Hanton, M. J.; Daubney, L.; Lebl, T.; Polas, S.; Smith, D. M.; Willemse, A., Dalton Trans, 2010, 39, 7025.
18 Gibson, V. C.; Spitzmesser, S. K., Chem. Rev., 2002, 103 , 283.
16 Chapter 1 Introduction
19 Agapie, T.; Schofer, S. J.; Labinger, J. A.; Bercaw, J. E., J.Am. Chem. Soc., 2004, 126 , 1304.
20 Tobisch, S., Organometallics, 2007, 26 , 6529.
21 Small, B. L.; Marcucci, A. J., Organometallics, 2001, 20 , 5738.
22 Yang, Q.-Z.; Kermagoret, A.; Agostinho, M.; Siri, O.; Braunstein, P., Organometallics, 2006, 25 , 5518.
23 (a) Agapie, T.; Schofer, S. J.; Labinger, J. A.; Bercaw, J. E., J. Am. Chem. Soc., 2004, 126 , 1304. (b) McGuinness, D. S., Organometallics, 2008, 28 , 244. (c) Overett, M. J.; Blann, K.; Bollmann, A.; Dixon, J. T.; Haasbroek, D.; Killian, E.; Maumela, H.; McGuinness, D. S.; Morgan, D. H., J. Am. Chem. Soc., 2005, 127 , 10723.
24 Tomov, A. K.; Gibson, V. C.; Britovsek, G. J. P.; Long, R. J.; van Meurs, M.; Jones, D. J.; Tellmann, K. P.; Chirinos, J. J., Organometallics, 2009, 28 , 7033.
25 Atkins, P. W., Introduction, The detection of energy levels: spectroscopy in Physical Chemistry Fifth Edition, Oxford University Press, Oxford, 1994, 10.
26 van der Veen, R. M.; Kas, J. J.; Milne, C. J.; Pham, V.-T.; Nahhas, A. E.; Lima, F. A.; Vithanage, D. A.; Rehr, J. J.; Abela, R.; Chergui, M., Phys. Chem. Chem. Phys., 2010, 12 , 5551.
27 (a) Newton, M. A.; Dent, A. J.; Fiddy, S. G.; Jyoti, B.; Evans, J., Catal. Today, 2007, 126 , 64. (b) Newton, M., Top. Catal., 2009, 52 , 1410.
28 Evans, J.; Tromp, M., J. Phys.-Condens. Matter, 2008, 20 , 184020.
29 Stoumltzel, J.; Luumltzenkirchen-Hecht, D.; Frahm, R., J. Synchrot. Radiat. 2011, 18 , 165.
30 Dent, A. J., Top. Catal., 2002, 18 , 27.
31 Grunwaldt, J. D.; Molenbroek, A. M.; Topsoe, N. Y.; Topsoe, H.; Clausen, B. S., J. Catal. 2000, 194 , 452.
32 Bressler, C.; Saes, M.; Chergui, M.; Grolimund, D.; Abela, R.; Pattison, P., J. Chem. Phys., 2002, 116 , 2955.
33 Chen, L. X.; Jager, W. J. H.; Jennings, G.; Gosztola, D. J.; Munkholm, A.; Hessler, J. P., Science, 2001, 292 , 262.
34 Singh, J.; Lamberti, C.; van Bokhoven, J. A., Chem. Soc. Rev. 2010, 39 , 4754.
35 Chermette, H., Coord. Chem. Rev. 1998, 178, 699.
17 Chapter 1 Introduction
36 Isao, T.; Hirohiko, A., J. Phys. D. App. Phys. 1996, 29 , 1725.
37 van der Veen, R. M.; Kas, J. J.; Milne, C. J.; Pham, V.-T.; Nahhas, A. E.; Lima, F. A.; Vithanage, D. A.; Rehr, J. J.; Abela, R.; Chergui, M., Phys. Chem. Chem. Phys., 2010, 12 , 5551.
18 Chapter 2 X-ray Techniques
2 X-ray Techniques 2.1 Introduction The primary focus of this project has been the development of resonant inelastic x-ray scattering (RIXS) as a tool to probe transition metal complexes, in addition to the related technique of high energy resolution fluorescence detected (HERFD) x-ray absorption near edge structure (XANES) spectroscopy, and the background to these techniques covered are discussed here.
One of the main objects of this project was to gain a more in depth understanding of the XANES region of an absorption spectrum for 4d and 5d metals. As it directly probes the unoccupied density of states, it has the potential to reveal a great amount of detail about the electronic structure of a metal, as well as its geometry. However without a well established quantitative method this has not been fully exploited in the past. Resonant Inelastic X-ray Scattering (RIXS) was the main technique used in this project, with the aim of gaining a detailed understanding of the electronic and geometric effects that contributed to a spectrum, so in future this spectroscopy can be successfully used in studying homogeneous catalysts in situ. RIXS is a powerful technique used to probe both the occupied and occupied local density of states, providing high spectral resolution and offering two-dimensional spectra to correlate features.
2.2 X-ray Absorption Spectroscopy 2.2.1 X-rays and the Photoelectric Effect When considering x-ray absorption spectroscopy (XAS) we must first consider the nature of x-rays and the photoelectric effect. X-rays are a form of electromagnetic radiation which have short wavelengths (compared to visible light), and hence high energy. An important factor in the understanding of x-rays is the concept of particle- wave duality; x-rays exist as quantised units of light called photons, but also exhibit wave-like properties 1.
The photoelectric effect describes a photon of sufficient energy hitting a material which frees a core electron and promotes it into the continuum. The x-ray will be absorbed and any excess energy above the binding energy of the promoted electron will be released in the form of a photon. The energy needed to free the electron will vary between different elements, and the different orbitals of these elements. Each binding energy is unique; therefore XAS is an element specific technique. After this absorption
19 Chapter 2 X-ray Techniques process the atom will be in an excited state, and have a core-hole. We will now look at the photoelectron produced in this process.
Continuum
Photoelectron
Photoelectron
X-ray X-ray
Figure 2.1 - Diagrams demonstrating the photoelectric effect. On the left an energy level diagram showing the electron promotion from a core shell to the continuum on absorption of an x-ray. On the right is a diagram showing the emission of a photoelectron and its potential interaction with neighbouring atoms.
The photoelectron released from this process will then interact with neighbouring atoms; giving rise to a process known as backscattering. The outgoing photoelectron waves will reflect off neighbouring atoms, resulting in backscattered waves. These will then interact with the outgoing photoelectrons. The interactions may result in constructive or destructive inference to the wave function. From analysing the patterns of the wave function in Extended X-ray Absorption Fine Structure (EXAFS) we can determine the distance, number and type of neighbouring atoms. The excited state that the atom is in after promotion of a core electron is inherently unstable so an electron from a higher molecular orbital will drop down to fill the core hole, emitting a photoelectron in the process.
2.2.2 Synchrotrons and Beamlines Although it is possible to carry out x-ray absorption experiments using a lab based source the vast majority of modern research is carried out using synchrotron light sources, which offer much higher quality spectra in much faster experiments. A synchrotron is a particle accelerator which is capable of producing very intense light, in the x-ray region of the electromagnetic spectrum (as well as infrared and ultraviolet regions). Originally they were used for particle physics experiments but it was noticed
20 Chapter 2 X-ray Techniques that they also produced high energy light and this was exploited for x-ray experiments in a parasitic fashion. Later, synchrotrons were built for the use of this light, so called 2nd generation synchrotrons, optimised so they can produce high energy electromagnetic radiation which is well adapted for their purpose 2. Furthermore, 3 rd generation synchrotrons were built from the mid 1990s, these use insertion devices, special sets of magnets, which provide brighter and more tuneable X-rays. A diagram showing the simplified layout of a synchrotron is shown below in Figure 2.2.
Storage Booster ring ring
Linac
Electron Beamline gun
Figure 2.2 – Schematic diagram of a synchrotron
The electron gun is used to generate electrons which are then channelled through the linear accelerator (linac) and booster ring (under ultra high vacuum) to accelerate them to high speeds (and therefore energies) before they enter the storage ring. In the storage ring there are bending magnets which guide the electrons around the ring, as well as straight connecting sections. The electrons, which are travelling at relativistic speeds, will lose energy when passing through the bending magnets, releasing light perpendicular to the direction of the electrons. This light is then guided into the beamlines, which lie at tangents to the storage ring. These allow further manipulation of the light to suit the purpose of the experimental stations, which lie at the end of the beamlines.
In modern, 3 rd generation synchrotrons the set-up of the synchrotrons also includes insertion devices, installed in the straight sections of the storage ring. A wiggler, also known as a wavelength shifter, is a multi-pole magnet which passes the electrons through chicanes 3 in order for them to follow a ‘wiggling’ path. This causes a broader spectrum of x-rays to be produced. The changing poles of the magnet cause the
21 Chapter 2 X-ray Techniques electron beam to follow a sinusoidal path 4. An undulator consists of an array of small magnets along the path of the electron beam which causes the beam to follow a wiggling path; each ‘wiggle’ causes x-rays to be formed in the direction of travel of the electron beam. There is constructive interference between the x-rays emitted from each ‘wiggle’ which produces a very intense beam in this direction 5.
In summary, thanks to modern insertion devices, synchrotron light has a broad range of wavelengths, from IR to high energy x-rays, with high flux, intensity and collimation. This radiation can be quasi continuous or pulsed and can be polarised 3.
Once the x-ray radiation leaves the storage ring it enters the beamlines, which are designed to optimise and control the radiation for the experiments taking place. There will be bending magnets and insertion devices, as previously discussed, in order to control the radiation. In order to optimise the x-rays for their end purpose there are a number of components in the optics hutch, which work to select the energy of the x- rays, to collimate the x-rays and to focus the beam. A simplified beamline set-up for XAS experiments is shown below in Figure 2.3.
Storage Ring Entrance Slits slits End Station Focusing mirror Aperture Double Exit and shielding crystal slits monochromator
Figure 2.3 - Diagram of a typical XAS beamline
Synchrotron light is polychromatic but a monochromatic beam is needed in order to control and vary the energy during an experiment. For this purpose a monochromator is used, which is based on Bragg crystal optics for hard x-rays. The incoming beam is diffracted off a crystal according to the Bragg equation:
nλ = 2d sin θ
Equation 2.1 - Bragg equation
22 Chapter 2 X-ray Techniques
Where n is an integer, λ is wavelength of the incoming x-rays, d is the lattice spacing of the crystals and θ is the angle between the incoming x-rays and the planes of the crystal. Typically θ is varied in order to control the energy of the incoming x-ray radiation that will be used in the experiments. Commonly a pair of crystals is used to increase the energy resolution. The monochromator crystals are typically made of silicon, for example pairs of Si(111) and Si(311) crystals were used in this project. The intrinsic energy resolutions of these crystals are 1.4 × 10 -4 DE/E for Si(111) and 0.3 x 10 -4 DE/E for Si(311) 6.
Coated mirrors (e.g. Pt, Rh) are used to focus the beam and to suppress higher harmonics and slits are used to control the size of the beam.
After the optics hutch we come to the experimental hutch. Here the different components are designed to create the optimum sample environment, as well as safety features such as interlocking. Here we may have UHV (ultra high vacuum) chambers, cryostats and/or motorised sample stages in which to mount specialised cells. There will also be detectors.
2.2.3 Detection methods There are several ways to detect the emitted photons from an x-ray spectroscopic experiment, the three most common will be discussed here. The most straightforward of these is transmission, where the incident x-ray intensity is compared to the transmitted intensity.
Reference Sample material
Incident I I I x-rays 0 1 2
Figure 2.4 – Detector set-up for a transmission XAS experiment
In Figure 2.4 the diagram shows the set-up of a typical transmission experiment. The x- ray intensity is measured before (I 0) passing through the sample and then again after
23 Chapter 2 X-ray Techniques
7 (I 1). They are typically measured using gas ionisation chambers , or photodiodes. We can then plot log(I 0/I 1) against energy for our XAS spectra. We can also measure the reference spectrum with a reference material and an additional detector (I 2) after the sample and the I 1 detector. This is helpful in calibration, which was previously important when beamlines were less stable.
Sometimes transmission may not be the optimal detection method. For example if the sample is too thin or too dilute then there will be little difference between the intensity before and after the sample, resulting in poor signal to noise ratio. In these cases it is then helpful to use fluorescence detection, which is more successful at measuring dilute samples. Another problem may be that if the samples are too thick the beam may be completely attenuated at the edge. A sample which is too concentrated needs to be diluted before measurement.
Fluorescence detection measures the emitted photons from the sample. It will be proportional to the absorption coefficient 8 as the photons are emitted when valence electrons drop down to fill core holes left by absorption. Full or partial fluorescence yield can be measured. Partial yield requires highly efficient detectors with good energy resolution in order to separate the emission from different species. For this solid-state germanium or drifted silicon detectors are typically used 57 . Fluorescence measurements are ideal for dilute samples because they are bulk sensitive and have high selectivity.
For concentrated samples there is often the issue of self absorption 9 which can dampen the features of the spectra and lower the absorption co-efficient. This is due to the reduced penetration depth with increasing absorption and the re-absorption of fluorescence photons by the same species within the same material 10 . However there are methods to correct this problem 9,10 .
The third commonly used detection method is Total Electron Yield (TEY). When x-rays are absorbed by a sample electrons are emitted proportional to the absorption coefficient 8. This can either be measured using electron multipliers, which monitor the electrical current of the sample, or by gas ionisation using a grid positioned close to the surface of the sample 11 . Whereas TEY collects all the electrons from this process an alternative method is Partial Electron Yield (PEY), where an energy analyser is used to differentiate between the photoelectrons and Auger electrons. Auger electrons are created when the a system is in its excited state following x-ray absorption, to fill the
24 Chapter 2 X-ray Techniques core hole an electron will drop down from a higher level. The energy emitted in this process can be emitted as fluorescence radiation, or non-radiatively by kicking out an electron 12,62 . This method is surface sensitive due to electron scattering and will only measure the first 10s of nm of the sample, which must be solid 13 . The sample needs to be conductive, or a thin layer on a conductive holder. Another problem with this detection method is that it is not possible to select which emission energy is measured, so it is therefore not useful for RIXS.
2.2.4 X-ray Absorption Fine Structure Spectroscopy X-ray Absorption Fine Structure (XAFS) encompasses both Extended X-ray Absorption Fine Structure (EXAFS) and X-ray Absorption Near Edge Structure (XANES), two commonly used techniques which will be discussed in this chapter.
The diagram below (Figure 2.5) shows a simplified set-up for a typical x-ray absorption experiment on a scanning beamline at a synchrotron.
Slits Ion chamber Ion chamber
Slits
X-ray Sample Reference beam Monochromator
Fluorescence detector
Figure 2.5 - Diagram of simplified set-up at scanning x-ray absorption spectroscopy beamline
The x-ray beam enters the beamline from the storage ring and is passed through a set of slits to control its size. It then passes through the monochromator, a pair of perfect parallel crystals which use Bragg optics to control the energy of the x-ray beam. There may be another set of slits to further control the size, and a set of mirrors to suppress higher harmonics. Once the beam is the required size and energy it will pass through an ion chamber, I 0, then through the sample before reaching another ion chamber, I t.
From the signals at I 0 and I t we can measure our sample in transmission. If the sample is particularly dilute we can measure using a fluorescence detector, shown here at the 90° from the beam to gain maximum signal and minimal scattering.
25 Chapter 2 X-ray Techniques
During the experiment the monochromator will scan through a range of energies, to produce a spectrum such as the one seen in Figure 2.6. As the energy increases the spectrum is initially featureless, until we reach 20 keV where we see an edge step. This is when the energy of the x-rays reaches the binding energy of a core electron and the electron is promoted, releasing a photoelectron. This is a molybdenum K-edge spectrum, so it is the promotion of a 1s electron.
EXAFS
Pre-edge and XANES
Figure 2.6 -X-ray absorption spectrum for Mo(CO) 6
The area around the edge, up until about 30-50 eV past the edge, is known as the X- ray Absorption Near Edge Structure (XANES) region. This gives information on the geometry and electronic structure around the absorber atom. This region corresponds to core-to-valence transitions. This is where the electron is excited into the lowest unoccupied molecular orbitals (LUMO).
From the end of the XANES region onwards we have the Extended X-ray Absorption Fine Structure (EXAFS) region, a series of oscillations. From this region structural information can be deduced, including the type and number of neighbours, the distance of neighbouring atoms and the disorder of the system. This region corresponds to core- to-continuum transitions, which generate photoelectrons which then cause backscattering 14 , discussed in the following section.
26 Chapter 2 X-ray Techniques
XAFS measures the x-ray absorption co-efficient µ(E), where E is the photon energy 16 , ρ is the sample density, Z is the atomic number and A the atomic mass.
ρZ 4 µ(E) ≈ Equation 2.2 – Absorption co-efficient AE 3 This describes how strongly x-rays are absorbed as a function of E. As the energy increases the absorption coefficient will decrease and the x-rays will be more penetrating. There is also a strong dependence on Z which explains why the absorbance of x-rays of the same energy varies between different elements.
The Beer-Lambert equation (Equation 3) describes the attenuation of the x-ray intensity, where I is the x-ray intensity, I 0 is the incident x-ray intensity and x is the path length.
−µ(E) x I = I0e Equation 2.3 - Beer-Lambert expression
The X-ray absorption coefficient, µ, governs the intensity of the absorption edge and is proportional to the probability of photon absorption as shown in the one electron approximation of Fermi’s golden rule 15 (Equation 4).
µ ∝ 〈ψ p ⋅ A r ψ 〉 2δ E − E − hω ∑ f ( ) i ( f i ) Equation 2.4 - Fermi's golden rule f
Ψi and Ψf are wavefunctions describing the initial and final states respectively, p is the momentum operator and A(r) is the vector potential of the incident electromagnetic field 16 .
A major advantage of XAFS is that it is element specific, because we tune the incoming X-rays to a particular absorption edge which will be unique to the element in question. However this can also lead to difficulties if that element is present in multiple forms within the material, the resulting spectrum will be an average of all these different forms. The most common edges to be studied are K edges and L edges. The K edge is the transition of a 1s electron to the lowest unoccupied p orbital. The L 2,3 edges are the transitions from 2p to a d orbital. Therefore the L3 edge directly probes the d density of states (DOS) of a metal 15 . The K-edge only probes this indirectly; in certain geometries
27 Chapter 2 X-ray Techniques we see s to d transitions as a result of hybridisation between p and d orbitals. The K- edge is at higher energies than the L-edges, and for 3d and 4d transition metals a vacuum environment is needed to measure the L-edges, therefore in-situ measurements are not generally possible.
As it is a spectroscopy XAFS can be used for disordered systems, making it suitable for both homogeneous and heterogeneous catalytic investigations 14 . One drawback to XAFS is that it is very difficult to distinguish between elements which are close together in the periodic table, i.e. with a similar Z. So for example C, O and N are hard to distinguish which can be problematic when studying organometallic systems.
2.2.5 Extended X-ray Absorption Fine Structure (EXAFS) Spectroscopy Extended X-ray Absorption Fine Structure (EXAFS) refers to the relatively weak oscillations which occur from around 20-30 eV past the edge in an x-ray absorption spectrum 16 , this is illustrated in Figure 2.6. These oscillations are caused by interference between the emitted photoelectron and neighbouring atoms, in a process known as backscattering. Thanks to many theoretical developments over the last century the EXAFS can be interpreted to give detailed information on bond lengths, neighbouring atom types and co-ordination numbers, as well as their disorder (both statistical and thermal).
The first fine structure around an absorption edge was observed by Fricke (1920) 17 when looking at the K-edges of compounds including metals such as iron and chromium. The first theory which accounted for the fine structure visible in x-ray absorption spectroscopy was published by Kronig (1931) 18 , describing long range order effects that often had poor agreement. He then published an alternative theory in 1932 19 which attributed the oscillations in EXAFS to neighbouring atoms, a short range order effect 16 .
Short-range-order theory describes the outgoing photoelectron from the absorber atom as having quantum mechanical wave-like behaviour 16 , neighbouring atoms reflect the wave back to the original atom. These reflected waves can interact constructively or destructively with the original photoelectron wave giving rise to oscillations. An important factor when considering the validity of short-range-order is the decay of an excited photoelectron over time. The outgoing photoelectron will be too weak to reflect waves off distant atoms when it has travelled too far from the original atom. The reflected waves will have the same problem and those travelling from more distant
28 Chapter 2 X-ray Techniques
atoms will be considerably weaker when returning to the original atom compared to direct neighbours. This means that the extent to which we can measure the local structure of a material is restricted by the effective mean free path of the outgoing photoelectron (i.e. net lifetime) 16 .
When considering EXAFS we often use k, the photoelectron wave vector. This describes the outgoing photoelectron as shown in equation 5, where m e is the mass of the photoelectron and E 0 is the energy at the edge.
8π 2 m k = e hν − E Equation 2.5 - Photoelectron wave vector 2 ()0 h
In 1971 Sayers, Stern and Lytle published their theory 20 which was able to calculate the distance and number of neighbouring atoms. This and further work over the next few years led to the EXAFS formula which is still used as standard 16 :
Equation 2.6 – EXAFS equation
2 f (k) 2 / ( ) 2 2 2 sin 2 2 − R λ k − σ k χ()k = ∑ S0 N R 2 ()kR + δ c + φ e e R kR
In this formula R represents the interatomic distances, N R is the co-ordination number, f(k) is the back scattering amplitude, this describes the reflection of the photoelectron waves which depends on the type and number of neighbouring atoms. λ(k) is the energy dependent mean free path, the exponential term which describes the decay of the photoelectron with distance, and1/kR 2 is the spherical wave factor.
δc is the phase shift at the absorbing atom, it describes the difference between the
2 / geometric and measured interatomic distances. e− R λ accounts for the decay of the photoelectron wave due to the mean free path, this includes the core hole lifetime. This
2 2 2 factor governs the short range of EXAFS. e− σ k is an approximation of the Debye- Waller factor which is partly due to thermal effects causing particles to move around. Φ is the phase factor which includes the quantum mechanical nature of the backscattering waves and the σ2 term factors in structural disorder.
29 Chapter 2 X-ray Techniques
2 S0 is the amplitude factor, and is a many-body effect due to the relaxation of the system when the core hole is created. It is usually approximated by a constant, despite having a weak energy dependence. Sin(2kr) describes the dependence of the oscillations on interatomic distances and energy.
This equation makes quantitative analysis of EXAFS possible, enabling us to obtain information about the structure around an absorbing atom by accurately interpreting the oscillations of the spectra. The analysis requires good, accurate scatterings, amplitudes and phase shifts.
Modern EXAFS analysis techniques have improved on this original equation in some respects. For example it was recognised that as well as backscattering (direct reflection back to absorber atom) there is also multiple scattering 16 . This refers to scattering paths where the electron wave is reflected off multiple atoms in succession before returning to the original atom.
The XANES region of an absorption spectrum requires different physical considerations than EXAFS 16 and a quantitative theory is still under development. This is discussed further in section 2.3.
2.3 X-ray Absorption Near Edge Structure (XANES) 2.3.1 Introduction The XANES region is the area around the absorption edge, up to about 50 eV past the edge, and provides information on the oxidation state and coordination of the metal. It results from the excitation of a core electron to the valence states. The structure of this near edge region is relatively little understood when compared to EXAFS, but it is due to numerous effects and it is thought that it mainly reflects the unoccupied density of states of the metal 62 . This allows us to look at effects such as hybridisation of bonding orbitals. In the past XANES has mainly been used as a fingerprinting technique to determine the oxidation state or geometry of a material, by using known reference materials 21 , but recently developments have been made in the effort to use XANES spectra to gain real information about the electronic structure, geometry and oxidation state of the metal (for example 22 ).
30 Chapter 2 X-ray Techniques
2.3.2 Multiplet effects Multiplet effects can have a significant effect on the shape of x-ray absorption spectra. They occur when there is a strong overlap of core and valence electron wavefunctions 23 in the intermediate or final states, i.e. in the presence of a core hole. For example in L-edge XANES of a 3d transition metal the final state of the process will have a partially filled core state, i.e. a 2p 5 configuration. If the 3d band is also partially filled, then the 2p and 3d holes will have wave functions that significantly overlap which can be large atomic effect 23 . There are much bigger multiplet effects for shallow core levels than the deeper core levels, for example in K-edge XANES multiplet effects are generally not visible, but they have a significant effect for L-edge spectra of 3d transition metals 23 . The multiplet effects are relatively small for L-edge spectra of 4d and 5d metals, there is relatively little overlap between the 2p and 4d or 5d wavefunctions compared to the analogous spectra for 3d metals 23 .
The multiplet effects for L-edge spectra of 3d metals have a strong influence on the fine structure, so analysing this can provide us with detailed information. Multiplet effects reflect the occupancy of the d levels so calculations using, for example, atomic multiplet theory give information on the oxidation state 30a.
2.3.3 Studies of K edge XANES There are numerous studies of XANES as a fingerprint technique in the literature. In this technique XANES spectra are taken of well defined systems with known geometries and oxidation states, before being compared to XANES spectra of unknown samples. This has often been carried out with K-edge XANES spectra of 3d transition metal compounds. In these systems the pre-edge region contains important information about the geometry so the shape of this area is compared, in addition to the edge shift, which partially corresponds to the oxidation state. Examples include the study of the Phillips catalyst by Groppo et al 21 where XANES spectra of chromium oxide references were measured and compared to spectra of the catalyst in order to identify the chromium species present. Although a detailed analysis was not performed XANES can still be used successfully in these experiments to deduce structural and electronic information.
Selection rules that determine the symmetry of the orbitals participating in the transition arise from the coupling of the electromagnetic field of the photon with the initial and final electronic states 54 . The XANES spectra will be broadened by the lifetime of the 1s
31 Chapter 2 X-ray Techniques core hole. For example the K-edge lifetime broadening for Cr is 1.08 eV, for Mo is 4.52 eV and for W is 33.9 eV 24 .
Although the majority of studies have used XANES as a simple fingerprinting technique the spectra contain a huge amount of structural and electronic information, and complexes with different formal oxidation states can still produce very similar XANES spectra 25 . There has been success in analysing the effect of geometry of 3d complexes from the pre-edge region, which often contains a white line feature. This white line is the result of 1s to 3d spin forbidden quadrupole (∆l=±2) transitions (for 3d metals), and therefore gives geometrical information. For this to occur there needs to be a degree of hybridisation between p and d orbitals, and this occurs to a greater degree for complexes of lower symmetry 26 . Therefore when the 1s to 4p transition occurs there will also be some intensity in the pre-edge due to 3d character in the transitions. For a complex with perfect octahedral symmetry hybridisation is not possible, but if distortion is introduced around the metal site then some intensity will be seen in the pre-edge 27 . The overall analysis of this region is complicated by the presence of multiplet effects 28 and the broadening, which usually results in a value of around 1.5eV for 3d metals for normal XAS experiments, due to a combination of the lifetime broadening and the experimental resolution 28 .
2.3.4 L edge XANES L-edge XANES has a big advantage over K-edge – it directly probes the d-orbitals, which govern the chemistry of the transition metals. For example compounds can be formed with a variety of oxidation states due to the relative stability of unpaired d electrons, and in turn this governs paramagnetism 29 . A partially filled d shell leads to coloured compounds due to d-d transitions. This direct d orbital information means that L-edge XAS is well suited to studying catalysts, as changes in the d electron configuration during reactions can be studied.
30 It is now possible to measure the L 2,3 edges of 3d metals thanks to the development of high resolution soft X-ray grating monochromators, and the L-edges of 4d and 5d metals are easily accessible using hard x-rays. In L-edge experiments the transition probed is from a 2s or 2p orbital to the lowest unoccupied d orbitals, the spectra reflecting the unoccupied d density of states. For a 5d transition metal L 1 is the transition from a 2s orbital to a 5d upon absorption of a photon, L 2 and L 3 are the transitions from 2p 1/2 and 2p 3/2 orbitals to empty 5d orbitals respectively. The lifetime broadening for all the L-edge transitions is relatively small due to the longer lifetimes of
32 Chapter 2 X-ray Techniques
the 2p coreholes compared to the 1s coreholes at the K edge. For example the Mo L 3- edge lifetime broadening is only 1.74 eV compared to 4.52 eV for the Mo K-edge.
For L-edge spectra of 3d transition metals multiplet effects are an important consideration due to the large coulombic overlap between the 2p corehole and the 3d electrons 31 , causing a big effect on the spectral shape. This accounts for the success in modelling L edge spectra for 3d metals using multiplet calculations 31,23 However for 4d and 5d systems the valence electrons are quite delocalised from the intermediate core hole, so multiplet effects do not have such a significant effect as for the 3d systems. However many L-edge experiments for the 3d and 4d transition metals are at low energies so require measurements to be taken in vacuum, therefore in-situ experiments would be extremely challenging.
There have been many studies on the L 2,3 edges of the first row transition metals, where the 2p to 3d transition is probed, which show intense features, detailed fine structure and show sensitivity to changes in oxidation state. From a detailed analysis these spectra have provided information on the ligand field and symmetry of the metal, the spin-state and covalency 23,32 .
2.3.5 Molybdenum XANES Molybdenum, as a 4d transition metal, has a range of experimentally accessible absorption edges. The Mo K-edge is found at around 20 keV, and the near edge structure is dominated by dipole forbidden 1s to 4d transitions, so the intensity of pre- edge features is proportional to the degree of Mo s, p and d hybridisation in the lowest 32 32 unoccupied orbital . The L 1 edge (~2866 eV) contains similar information . The Mo
L2,3 edges (2625 eV and 2520 eV respectively) probe the dipole allowed 2p to 4d transitions and have been shown to be sensitive to the ligand field and oxidation state of the Mo atom, in studies on a number of systems including catalysts 39 and enzymes 33 . These experiments all require the samples to be measured in a vacuum due to the low energy of the x-rays.
It is also possible to measure the M edges of molybdenum, although this is experimentally more challenging due to the low energies, around 400 eV for M 2,3 (3p to 33 4d excitation) and around 230 eV for M 4,5 (3d to 4d) As soft x-rays have short path- lengths an ultra-high vacuum set-up is required. Additionally there is the possibility of the Mo M edges being obscured by light elements (e.g. C, N, O) with K-edges in the same region 33 . The low fluorescence yield at these edges 33 also presents difficulties,
33 Chapter 2 X-ray Techniques necessitating the use of modern high-resolution superconducting tunnel junction (STJ) detectors 33 that able the measurement of XANES for these edges. Due to these difficulties few studies exist but it is thought that Mo M-edges have a low natural linewidth 33,34 and they could be useful probes of electronic structure and oxidation states 33 . In this work we will study the M-edge indirectly using 2p4d Mo RIXS, which is discussed in chapter 4.
There are a number of existing direct L-edge molybdenum XANES studies in the literature, particularly for the metal oxides 35,36 , which will provide an interesting comparison to the Mo L α RIXS carried out in this work. They all show a characteristic “white line” due to the dipole allowed 2p to 4d transition, and the splitting of this peak can be directly related back to the geometry (for simple geometries) 37,38,39 .
Figure 2.7 - Normalised fluorescence yield XANES taken from Hu and Wachs 39 . Reprinted with permission from J. Phys. Chem., 1995 , 99, 10901. Copyright 1995 American Chemical Society.
Figure 2.7 shows examples of molybdenum L 3 XANES spectra and it can be seen that for the sodium molybdate, with the central Mo atom in a tetrahedral environment, the intensity ratio is approximately 2 : 3, whereas for the molybdenum (VI) oxide (octahedral) the intensity ratio is reversed and the first peak is of greater intensity. This is consistent with similar studies in the literature. This peak results from the 2p to 4d
34 Chapter 2 X-ray Techniques transition so by looking at the relative intensities of the splitting we can relate this to the d orbital ligand field splitting (Figure 2.8).
dz2 ,d x2-y2 Energy
∆oct Octahedral
dxy , d xz , d yz
Energy dxy , d xz , d yz
∆tet Tetrahedral
dz2 , d x2-y2
Figure 2. 8 – Energy level diagram showing d-orbital crystal field splitting for octahedral and tetrahedral environments
2.3.6 Tungsten and rhenium XANES For tungsten and rhenium the K-edges are too high to measure experimentally (~70 keV), so L-edges are the main route of study. Tungsten L-edge XANES has been a useful technique in heterogeneous catalysis, to study the local geometry of tungsten species on supports. The L 1 edge of tungsten oxides has been used to successfully predict the local geometry in the study by Horsley et al (1987) 40 which explores the different geometries and distortions present on the surface of WO 3/Al 2O3, showing the presence of coordinated water by the change from a tetrahedral to distorted octahedral geometry. In a similar study the L 1 and L 3 edges were used to probe the speciation of tungsten oxide species on titania and alumina supports in the work of Hilbrig et al 41 (1991) by looking at the pre-edge features. In the L 1 (2s to 6p) spectra a sharp and intense pre-edge peak is present when the W has tetrahedral geometry. These pre- edge peaks are much less sharp and intense as the geometries move towards the 41 distorted octahedral of WO 3 . These pre-edge features are due to 2s-5d transitions, made possible by the mixing of p orbitals from the oxygen atoms with the tungsten d orbitals 41 , and W p-d hybridisation. This is dipole forbidden in a perfect octahedral, but in distorted octahedral such as WO 3 the centre of inversion is no longer present, allowing weak transitions 41 . Therefore tetrahedral and octahedral geometry are easily identifiable using L 1 XANES spectra, as in K-edge XAS. As the L 3 edge probes the unoccupied d density of states we see d-band splitting, although not highly resolved in such studies 41 .
35 Chapter 2 X-ray Techniques
Similar examples exist for the use of rhenium L-edge XANES in the study of catalysis, 42 for example Re promotion of Co/Al 2O3 in Fischer-Tropsch synthesis and PtRe/Al 2O3 43 which is used in the commercial reforming of naptha . The L 3 XANES spectra probed in such studies show a similar sensitivity to the local geometry as the tungsten studies.
2.3.7 Energy resolution of XANES Due to the broadening effects of core holes, features in the XANES can be obscured, so higher resolution spectra are needed in order to see these, this is particularly important for catalysis when subtle changes or shifts may occur during the reaction, reflecting changes in oxidation state or geometry of the catalyst. The values relevant to the studies in this thesis are shown in Table 2.1.
Element Edge Core hole Core hole broadening/ eV
Molybdenum L3 2p 3/2 1.74
Molybdenum M4 3d 3/2 0.16
M5 3d 5/2 0.14
Tungsten L3 2p 3/2 4.98
Rhenium L3 2p 3/2 5.04
Table 2.1- Selected core hole broadenings 24
2.3.8 High Energy Resolution Fluorescence Detected XANES In a typical XANES experiment the unoccupied electronic density of states can be probed by scanning the incident energy around the absorption edge. One way in which this can be measured is with a fluorescence detector, for example solid state Ge, which gives a higher energy resolution than photodiode detectors 15 . However this energy resolution is partially limited by the life-time of the core hole. This problem has been partially solved by the technique known as high energy resolution fluorescence detected (HERFD) XANES 44 . In a HERFD experiment only the emitted photons from a particular fluorescence line are monitored. This decay pathway will have a shorter core- hole lifetime than the initial absorption process, and therefore less broadening, leading to XANES spectra with higher resolution and much sharper features 15 . For example the s core hole in a K-edge experiment can be replaced by a p or d core hole which has a much smaller lifetime broadening. The experimental technique is discussed in section
36 Chapter 2 X-ray Techniques
2.5, HERFD can be thought of as single line plots (constant emission energy) from RIXS planes.
It is possible to simulate the features of the HERFD structure using modern theoretical programs, such as FEFF 44 and the multiplet theory 45 which give important information on the electronic orbital contributions.
There have been a number of studies using HERFD to study catalytic reactions 46 , it has been shown that L-edge XANES is sensitive to the adsorption of molecules on, for example, Pt particles 47 and using HERFD to greatly reduce the lifetime broadening enabled the more subtle differences between adsorption sites and modes on Pt particles to be observed and interpreted.
2.4 X-ray Emission Spectroscopy X-ray emission spectroscopy (XES) and x-ray absorption spectroscopy (XAS) are closely linked. Whereas XAS probes the unoccupied density of states, XES probes the occupied electronic density of states 48 . Both are sensitive to the local electronic structure and geometry of the absorbing atom 54 . XES has a wide range of applications across many different areas of science including mineralogy 49 , the study of photosynthetic mechanisms 50 and fuel cells 51 .
After absorption of a photon and subsequent excitation of an electron to a higher molecular orbital there will be a core hole, to which an electron from a higher orbital will drop down to fill. In this process it will emit a photon, monitoring these emitted photons leads to emission spectra. Some transitions are more likely than others, and each occurs at a specific energy. XES is also, like XAS, bulk sensitive due to the penetration depth of hard x-rays, which is an advantage compared to similar techniques such as inner-shell x-ray photoelectron spectroscopy (XPS) which are limited to the surface. XES can measure the same final state configuration as XPS but has much more flexibility with regards to the sample environment, allowing different temperatures and pressures, ideal for in-situ/operando catalytic studies 52 .
X-ray emission spectra can be dependent on the incident energy, i.e. resonant, or independent of the incident energy, i.e. non-resonant XES. For non-resonant XES experiments it is not necessary to control the incoming x-rays, i.e. a monochromator is not needed, neither is a tuneable x-ray source 54 . The 1s electron is excited well above
37 Chapter 2 X-ray Techniques the edge and the resulting spectra do not significantly change with incident energy. However with resonant XES synchrotron radiation is tuned to the area around the absorption edge and the resultant x-ray emission spectra show a clear dependence on the incident energy.
Although XES does not always require high tuneability or resolution of the incoming x- rays, one important factor is having an analyser able to measure the emitted x-rays. The photon emitted during the decay is detected using a crystal analyser (as seen in Figure 2.9) with an energy bandwidth of the same order as the lifetime broadening caused by the core hole. The analyser is able to select the energy of the emission spectrum and this specific fluorescence is reflected onto a detector. This is therefore a second order process 53 and allows the study of the inelastic scattering of the incident photon at a particular resonance energy. The number of analyser crystals can be varied, one being the typical number but multiple crystals are useful for systems where the fluorescence may contain weak features; they provide a larger solid angle of collection. The multiple array also allows for improved time resolution which is particularly important for catalytic studies.
38 Chapter 2 X-ray Techniques
Figure 2.9 - Set-up of an x-ray emission multi-crystal spectrometer. Reprinted with permission of Glatzel, P.; Sikora, M.; Smolentsev, G.; Fernández-García, M., Catalysis Today 2009, 145 (3-4), 294.
M (3d ) 5 5/2 M4 (3d 3/2 )
M3 (3p 3/2 )
M2 (3p 1/2 )
M1 (3s)
Lα1 Lα2 Lβ1
L (2p ) 3 3/2 L2 (2p 1/2 )
Kα Kα Kβ L1 (2s) 1 2 1 K (1s)
Figure 2.10 - Selected emission lines
Figure 2.10 shows an energy level schematic with a selection of emission lines, to illustrate the terminology. The emission line first takes its name from the excitation process which creates the core hole. So an excitation from a 1s orbital leads to K
39 Chapter 2 X-ray Techniques fluorescence, excitation from a 2s or 2p orbital leads to L fluorescence. However there are different emission paths possible due to the occupation of various higher molecular orbitals. So these different fluorescence lines are labelled according to intensity, starting with α for the most intense, then β, γ etc 54 . So the most intense emission from an L-edge excitation is the L α1 emission line, which corresponds to the filling of a 2p 3/2 corehole with a 3d 5/2 electron. As with XAS it is also necessary to consider the selection rules which determine which orbitals may participate in a transition 54 .
Non-resonant XES can be carried out for any measurable emission line. The example discussed here is K β which has been used extensively for 3d transition metals. The main 3p to 1s K β main line corresponds to core-to-core transitions, whereas the K β satellite lines corresponds to valence-to-core 62 . The position of the K β main peak directly corresponds to the oxidation state of the metal, with less dependence on structure than XANES 54 . The satellite lines correspond to transitions from valence orbitals to the 1s shell and are chemically sensitive 44 , providing information on the electron orbitals that are involved in the chemical bonds. These emission lines are also sensitive to the ligand orbitals which aids in the identification of ligands bonded to the transition metal 55 .
2.5 Resonant Inelastic X-ray Scattering 2.5.1 Introduction As discussed in section 2.4, resonant x-ray emission spectra are dependent on the incident energy. There are two main types of resonant scattering processes, elastic and inelastic. Elastic scattering is when an incident photon interacts with electrons, causing a photon of the same energy to be emitted 56 . However we are looking at inelastic scattering, where the emitted photon is generally of lower energy than the incident photon 56 . The energy difference between these photons is transferred to the sample by causing electronic transitions, as in x-ray absorption 56 . Experimental set-ups are detailed in section 3.6.
Resonant inelastic x-ray scattering (RIXS) looks at the unoccupied and occupied density of states whilst offering high spectral resolution 53,57 . It studies the same region as x-ray absorption near edge structure (XANES) spectroscopy, but over a range of emission energies in order to uncover additional features. Both the incident and emitted photons are measured. It has been particularly successful with 3d transition metal complexes 58 , where it has successfully suppressed the 1s lifetime broadening for K
40 Chapter 2 X-ray Techniques edge spectra in order to see much greater detail in the spectral shape 59 , but there are few studies on 4d or 5d metal systems 57,60 . There is limited understanding of RIXS for 3d metals, despite the relatively numerous experimental studies.
1s2p RIXS has been used to obtain L-edge type information for 3d transition metals 28 . By scanning along the K pre-edge and edge region for the initial excitation. The difference between the initial and final state will be equivalent to the transition for L-edge XAS, but from a two-step process 62 . This makes RIXS an important technique, it enables users to gain L-edge information using hard x-rays, therefore not requiring a UHV set-up. This makes a wider range of experimental conditions accessible, such as high temperature and pressure, and enables in-situ catalytic studies 15 . L-edge information is potentially very interesting for these types of systems as it directly probes the valence d-electrons. RIXS also offers the potential for selective XAS, for example selectively probing the spin or oxidation state.
RIXS has been made possible by modern light sources (third generation synchrotrons), with intense beams thanks to undulator radiation, and the design of efficient x-ray emission spectrometers. The combination of these at facilities with high levels of user support have made high energy RIXS available to scientists across a variety of fields, including; geology, catalysis and environmental science 61 .
2.5.2 The RIXS process RIXS is a combination of XAS and XES. In a RIXS experiment the fluorescence of the emission process is measured as a function of the incident energy. We scan across an absorption edge, then detect the fluorescence from the decay of resonantly excited states in an energy dispersive manner 15,62 . In a RIXS experiment a secondary spectrometer with an energy bandwidth similar to the lifetime broadening is used to control the emission energy 62 . This allows the superior resolution seen in RIXS compared to normal XANES. For example in the 1s2p RIXS study of iron oxides the overall resolution in the RIXS was ~1 eV, compared to ~1.5 eV for K-edge XANES 28 . These values take into account the applied experimental resolution as well as the lifetime broadenings.
41 Chapter 2 X-ray Techniques
Ground State Intermediate State Final State
2p 54d 10 5d n+1
4d 2p Decay hννν= ωωω 2p 5d 2p 64d 95d n+1 Absorption Energy hννν= ΩΩΩ Transfer ΩΩΩ- ωωω
2p 64d 10 5d n
Figure 2.11 - Energy level scheme showing the 2p4d RIXS process
The RIXS process for 2p4d RIXS is shown in Figure 2.11, which is used for the molybdenum RIXS in Chapter 4. An atom in its ground state electronic configuration absorbs a photon equal to the binding energy of a 2p electron. This is then excited to the intermediate state, which has a 2p core hole. To fill this core hole an electron from a higher orbital, in this case 4d, will drop down to fill the core hole, resulting in a final state with a 4d hole. The lifetime broadening of this is less than the intermediate state (equivalent to the XAS broadening), so the broadening will be due to the 4d core hole rather than the 2p. The energy transfer is the difference between the initial and final state, also the difference between the incident photon energy and the fluorescence energy 15 .
Figure 2.12 - L α RIXS 2D contour plot for Na 2MoO 4
42 Chapter 2 X-ray Techniques
RIXS spectra are typically presented as 2D contour plots, with energy transfer plotted against incident energy, an example of an Lα RIXS plot for Na 2MoO 4 is shown below in
Figure 2.12. For this plot the incident energy range covers the area around the L 3 absorption edge, with the emission range covering the M4 and M5 emission lines. Red denotes a region of high intensity, green low intensity. The 2p 3/2 core hole broadening only affects the RIXS spectra along the incident energy direction 15,63.
From a RIXS plot we can extract a series of line plots, shown in Figure 2.13. Firstly we see the constant energy transfer (CET) plot in the top left; this is equivalent to keeping the final state constant whilst scanning across all the possible intermediate states. In the top right we see the constant emission energy (CEE) plot, which corresponds to the HERFD and is the diagonal cut through the main peaks of the RIXS plot. This is overlaid with the normal absorption plot, and we can clearly see the increased resolution in the HERFD. The two main peaks are more clearly separated and a new feature becomes visible at 4969 eV. The final plot is the constant incident energy (CIE), which keeps the intermediate state constant whilst the final states are scanned; i.e. an emission spectrum is obtained, at a specific incident energy.
Figure 2.13 - RIXS plot with extracted line plots. 64
43 Chapter 2 X-ray Techniques
2.5.3 Valence Band RIXS Valence Band RIXS is a relatively new area of study, mainly due to the improvements of modern synchrotrons providing more intense beams to study such a weak transition. VB RIXS looks at charge-neutral transitions in the valence shell, which allows us to study the electronic structure around the Fermi level 60 . One study by Glatzel et al 60 (2010) presents a series of Pt VB RIXS spectra collected at the Pt L 3 edge, two of which are seen in Figure 2.14. All the plots have a horizontal line of intensity at zero energy transfer, which will be referred to as the elastic peak. This is from transitions where there is no net energy gain for the sample, so can act as an absolute energy calibration. This allows us to mostly eliminate experimental errors such as energy drifts, therefore facilitating direct comparison between samples 66 .
Figure 2.14 - Pt L 3 VB RIXS plots for Pt nanoparticles (left) and Pt nanoparticles following the adsorption of CO (right) 65 .
The peaks we see in the plots at only a few eV energy transfer are the result of valence band excitations. As RIXS is a two-photon process we may see features that are optically forbidden in UV 60 . For the platinum nanoparticles seen in Figure 2.14 we see that the elastic peak and the valence band excitation peak are merged, this indicates that the there is a metallic structure and the Fermi level must lie within a partially filled band 60 . The Fermi level is at the position of the elastic peak (e.g. zero energy transfer). However when moving to the VB RIXS plot of the Pt nanoparticles with a CO adsorbate we can see there is now a clear gap between the elastic peak and the valence band excitation as the excitation is around 4 eV higher in the energy transfer direction. It is suggested that the valence orbitals are at deeper binding energies as a result of Pt-CO hybridisation 60 .
44 Chapter 2 X-ray Techniques
A recent study by Smolentsev et al (2011) 66 used core-to-valence RIXS (i.e. VB RIXS) to probe the occupied and unoccupied density of states of a set of four tungsten and rhenium oxides and are able to reproduce the spectra using ab initio single electron band structure calculations. This will be further discussed in Chapters 5 and 6.
2.6 Data Analysis 2.6.1 Summary In this project we will be analysing L α RIXS spectra for molybdenum compounds, HERFD XANES spectra for tungsten and rhenium compounds, and VB RIXS spectra, again for tungsten and rhenium compounds. The methods for simulating the data needed careful consideration due to the complex effects that contribute to a XANES or RIXS spectrum.
There is a range of different theoretical approaches that can be used to analyse x-ray absorption and emission spectra. Each theory has its own advantages and drawbacks, and no theory encompasses all the effects that contribute to an XAS or XES spectrum. It is therefore important to choose a method which is well suited to the systems being studied. For example when studying the L edge absorption of a 3d transition metal complex the charge transfer multiplet approach is a useful option as it takes into account the multiplet interactions between the valence electrons and core hole which is such a dominant effect here. However these multiplet effects have a much lesser influence when you move to 5d transition metal complexes as the valence electrons are much more delocalised and there will not be a significant overlap with the corehole wavefunction as with the 3d metals 23 . Due to these considerations programs based on the multiplet theory were not used in the analysis of the XANES and RIXS in this thesis.
Other common approaches to obtain a quantitative XANES analysis include band structure codes such as WIEN2K and PARATEC, ‘molecular’ DFT codes such as STOBE and ORCA and real space multiple scattering codes such as FEFF. Whilst most XAFS calculation use muffin tin geometry 16 in their treatment of atomic potentials DFT models are able to offer more sophisticated models. Full potential XANES calculations have proved successful for several types of systems 67 . However DFT models do not typically allow for core holes, an essential factor in RIXS spectra.
45 Chapter 2 X-ray Techniques
After some preliminary tests FEFF was chosen as an appropriate choice for calculating the XANES and RIXS spectra of 4d and 5d transition metal complexes used in this project and is explored below in more detail. One of the major reasons for choosing FEFF was that the entire crystal structure is used in the calculation input, this is important as we want to see the effects of structural changes in catalysis. This project involves the ongoing development of FEFF for use in these types of systems in collaboration with J.J. Rehr and J. Kas (University of Washington).
2.6.2 FEFF9 FEFF9 68 is able to quantitatively calculate spectroscopic properties for a range of techniques, this includes extended x-ray absorption fine structure (EXAFS), x-ray absorption near edge structure (XANES), x-ray magnetic circular dichroism (XMCD) and non-resonant x-ray emission spectroscopy (XES). It is also able to calculate the electronic structure including the local densities of states (LDOS). FEFF9 uses an ab-initio self-consistent real space multiple scattering approach based on spherical muffin-tin scattering potentials69 . It includes polarisation dependence, vibrational effects, core-hole effects and local field corrections. The calculations are based on a relativistic Green's function formalism.
In the FEFF calculations for XANES spectra full multiple scattering is calculated for clusters of atoms for both the XAS and the density of states (DOS). Within the calculation there are self consistent field (SCF) loops which calculate a number of parameters including the Coulomb potentials, the electron density, the Fermi energy, occupation numbers and the charge transfer between atoms 70 . The charge transfer will have a large effect on the chemical properties of the atoms, including their oxidation state and electron affinity; therefore it is a very important property in co-ordination compounds where metal-ligand bonding can include a significant amount of charge transfer 70 . Therefore we are able to obtain structural and electronic information from spectra based on SCF calculations, as they contain information on factors such as charge transfer and hybridization 70 .
The RIXS component of this thesis uses a new theoretical treatment of RIXS based on a real-space multiple-scattering Green’s function formalism, and includes the effects of both intermediate and final state core-hole interactions 68 . This exists as an extension to the current FEFF9 program which is still in the test phase. Additional many-body effects can be included via a convolution of the single particle spectrum with an effective spectral function, however in the present work, we have neglected these
46 Chapter 2 X-ray Techniques effects. This new method allows us to simulate Mo Lα RIXS spectra and W and Re VB RIXS spectra successfully for the first time using a convolution of an effective XAS spectrum and the XES spectrum of the complexes simply based on their crystal structures.
2.7 References
1 "Louis de Broglie - Nobel Lecture: The Wave Nature of the Electron". Nobelprize.org. http://www.nobelprize.org/nobel_prizes/physics/laureates/1929/broglie- lecture.html Accessed 29 Nov 2012
2 http://diamond.ac.uk/Home/About/Synchrotrons/history.html, About Diamond: A Brief History of Synchrotrons, Accessed 11/11/2012
3 Evans, J., Phys. Chem. Chem. Phys, 2006, 8, 3045.
4 Kunz, C., J. Phys- Condens. Mat. , 2001, 13, 7499.
5 http://insidediamond.org/undulator/ Accessed 11/11/2012
6http://www.esrf.eu/UsersAndScience/Experiments/DynExtrCond/ID26/Characteristic s ID26 Source Characteristics, Accessed 01/12/12
7 Pettifer, R. F.; Borowski, M.; Loeffen, P. W., J. Synchrot. Radiat. 1999, 6, 217.
8 X-ray absorption. Edited by D. C. Koningsberger and R. Prins. Published by John Wiley and Sons, New York, 1988
9 Pfalzer, P.; Urbach, J. P.; Klemm, M.; Horn, S.; den Boer, M. L.; Frenkel, A. I.; Kirkland, J. P., Phys. Rev. B, 1999, 60 , 9335.
10 Tröger, L., Arvanitis, D., Baberschke, K., Michaelis, H., Grimm, U., Zschech. E., Phys. Rev. B, 1992, 46 , 3283.
11 Kasrai, M., Lennard, W. N., Brunner, R. W., Bancroft, G. M., Bardwell, J. A., Tan, K. H., App. Surf. Sci. , 1996, 99 , 303.
12 de Groot, F.M.F. and A. Kotani, Fundamental Aspects of Core Level Spectroscopies, in Core Level Spectroscopy of Solids, 2008, CRC Press, 11-37.
13 Abbate, M., Goedkoop, J. B., de Groot, F. M. F., Grioni, M., Fuggle, J. C., Hofmann, S., Petersen, H., Sacchi, M., Surf. Interface Anal., 1992, 18, 65.
14 Newton, M. A.; Dent, A. J.; Evans, J., Chem. Soc. Rev., 2002, 31 , 83.
47 Chapter 2 X-ray Techniques
15 Singh, J.; Lamberti, C.; van Bokhoven, J. A., Chem. Soc. Rev., 2010, 39 , 4754.
16 Rehr, J. J.; Albers, R. C., Rev. Mod. Phys., 2000, 72 , 621.
17 Fricke, H., Phys. Rev. 1920, 16 , 202.
18 Kronig, R. d. L., Z. Phys. 1931, 70 , 317.
19 Kronig, R. D. L., Z. Phys. 1932, 75 , 191.
20 Sayers, D. E.; Stern, E. A.; Lytle, F. W., Phys. Rev. Lett. 1971, 27 , 1204.
21 Groppo, E.; Prestipino, C.; Cesano, F.; Bonino, F.; Bordiga, S.; Lamberti, C.; Thüne, P. C.; Niemantsverdriet, J. W.; Zecchina, A., J. Catal. 2005, 230 , 98.
22 van der Veen, R. M.; Kas, J. J.; Milne, C. J.; Pham, V.-T.; Nahhas, A. E.; Lima, F. A.; Vithanage, D. A.; Rehr, J. J.; Abela, R.; Chergui, M., Phys. Chem. Chem. Phys. 2010, 12 , 5551.
23 De Groot, F., Coord. Chem. Rev. 2005, 249 , 31.
24 Fuggle, J. C.; Inglesfield, J. E.; Andrews, P. J., Unoccupied Electronic States: Fundamentals for Xanes, Eels, Ips and Bis . Springer-Verlag: 1992.
25 Tromp, M., X-Ray Absorption Fine Structure-Xafs13, B. Hedman and P. Painetta, Editors. 2007, 699.
26 Frommer, J.; Nachtegaal, M.; Czekaj, I.; Weng, T.-C.; Kretzschmar, R., J. Phys. Chem. A, 2009 , 113, 12171.
27 de Groot, F., Chem. Rev., 2001 , 101,1779.
28 de Groot, F.M.F., Glatzel, P., Bergmann, U., van Aken, P. A., Barrea, R. A., Klemme, S., J. Phys. Chem. B, 2005 ,109 , 20751.
29 Matsumoto, P.S., J. Chem. Ed., 2005 . 82,1660.
30 (a)de Groot, F. M. F.; Fuggle, J. C.; Thole, B. T.;Sawatzky, G. A., Phys. Rev. B: Condens. Matter 1990, 41 , 928. (b) Cramer, S. P.; de Groot, F. M. F.; Ma, Y.; Chen,
C. T.; Sette, F.; Kipke, C. A.; Eichhorn, D. M.; Chan, M. K.; Armstrong, W. H., J.
Am. Chem. Soc. 1991, 113 , 7937.
31 de Groot, F.M.F. and A. Kotani, Core Level Spectroscopy of Solids, 2008, CRC Press, 93-144.
32 Funk, T.; Deb, A.; George, S. J.; Wang, H.; Cramer, S. P., Coord. Chem. Rev. 2005, 249 (1–2), 3-30.
48 Chapter 2 X-ray Techniques
33 George, G. N.; Cleland, W. E.; Enemark, J. H.; Smith, B. E.; Kipke, C. A.; Roberts, S. A.; Cramer, S. P., J. Am. Chem. Soc., 1990, 112 , 2541.
34 Campbell, J. L.; Papp, T., Atomic Data and Nuclear Data Tables, 2001, 77 , 1.
35 George, S. J.; Drury, O. B.; Fu, J.; Friedrich, S.; Doonan, C. J.; George, G. N.; White, J. M.; Young, C. G.; Cramer, S. P., J. Inorg. Biochem., 2009 , 103 , 157.
36 Bare, S. R.; Mitchell, G. E.; Maj, J. J.; Vrieland, G. E.; Gland, J. L., J. Phys. Chem., 1993 , 97 , 6048.
37 Evans, J.; Mosselmans, J. F. W., J. Phys. Chem., 1991 , 95 , 9673.
38 Aritani, H.; Tanaka, T.; Funabiki, T.; Yoshida, S.; Eda, K.; Sotani, N.; Kudo, M.; Hasegawa, S., J. Phys. Chem., 1996 , 100 , 19495.
39 Hu, H.; Wachs, I. E.; Bare, S. R. J. Phys. Chem. 1995 , 99 , 10897
40 Horsley, J. A.; Wachs, I. E.; Brown, J. M.; Via, G. H.; Hardcastle, F. D., J. Phys. Chem., 1987, 91 , 4014.
41 Hilbrig, F.; Gobel, H. E.; Knozinger, H.; Schmelz, H.; Lengeler, B., J. Phys. Chem. 1991, 95 , 6973.
42 Jacobs, G.; Chaney, J. A.; Patterson, P. M.; Das, T. K.; Davis, B. H., Appl. Catal. A-Gen. 2004, 264 , 203.
43 Hilbrig, F.; Michel, C.; Haller, G. L., J. Phys. Chem. 1992, 96 , 9893.
44 Safonova, O. V.; Tromp, M.; van Bokhoven, J. A.; de Groot, F. M. F.; Evans, J.; Glatzel, P., J. Phys. Chem. B 2006, 110 , 16162.
45 de Groot, F. M. F.; Krisch, M. H.; Vogel, J., Phys. Rev. B, 2002, 66 , 195112.
46 Safonova, O. V.; Tromp, M.; van Bokhoven, J. A.; de Groot, F. M. F.; Evans, J.; Glatzel, P., J. Phys. Chem. B 2006, 110 , 16162.
47 Oudenhuijzen, M. K.; van Bokhoven, J. A.; Miller, J. T.; Ramaker, D. E.; Koningsberger, D. C., J. Am. Chem. Soc., 2005, 127 , 1530.
48 Nilsson, A., Phys. Rev. B, 1996. 54, 2917.
49 Bargar, J. R.; Tebo, B. M.; Bergmann, U.; Webb, S. M.; Glatzel, P.; Chiu, V. Q.; Villalobos, M., Am. Miner. 2005, 90 , 143.
49 Chapter 2 X-ray Techniques
50 Messinger, J.; Robblee, J. H.; Bergmann, U.; Fernandez, C.; Glatzel, P.; Visser, H.; Cinco, R. M.; McFarlane, K. L.; Bellacchio, E.; Pizarro, S. A.; Cramer, S. P.; Sauer, K.; Klein, M. P.; Yachandra, V. K., J. Am. Chem. Soc. 2001, 123 , 7804.
51 Sun, J.; Qiu, X. P.; Wu, F.; Zhu, W. T., Int. J. Hydrog. Energy, 2005, 30 , 437.
52 Eeckhout, S. G.; Safonova, O. V.; Smolentsev, G.; Biasioli, M.; Safonov, V. A.; Vykhodtseva, L. N.; Sikora, M.; Glatzel, P., J. Anal. Atomic Spec. 2009, 24 , 215.
53 Kotani, A.; Shin, S., Rev. Modern Phys. , 2001, 73, 203.
54 Bergmann, U.; Glatzel, P., Photosynth Res 2009, 102 , 255.
55 Smolentsev, G.; Soldatov, A. V.; Messinger, J.; Merz, K.; Weyhermüller, T.;
Bergmann, U.; Pushkar, Y.; Yano, J.; Yachandra, V. K.; Glatzel, P., J. Am. Chem.
Soc. 2009, 131, 13161.
56 de Groot, F. M. F.; Kotani, A., In Core Level Spectroscopy of Solids , CRC Press:
2008; 335.
57 Glatzel, P.; Sikora, M.; Smolentsev, G.; Fernández-García, M., Catal. Today , 2009, 145 , 294.
58 de Groot, F. M. F.; Glatzel, P.; Bergmann, U.; van Aken, P. A.; Barrea, R. A.; Klemme, S.; Havecker, M.; Knop-Gericke, A.; Heijboer, W. M.; Weckhuysen, B. M., J. Phys. Chem. B, 2005, 109 , 20751.
59 Kotani, A.; Matsubara, M.; Uozumi, T.; Ghiringhelli, G.; Fracassi, F.; Dallera, C.; Tagliaferri, A.; Brookes, N. B.; Braicovich, L., Rad. Phys. Chem., 2006 , 75 , 1670.
60 Glatzel, P.; Singh, J.; Kvashnina, K. O.; van Bokhoven, J. A., J. Am. Chem. Soc. 2010, 132 , 2555.
61 Glatzel, P.; Bergmann, U.; Yano, J.; Visser, H.; Robblee, J. H.; Gu, W. W.; de Groot, F. M. F.; Christou, G.; Pecoraro, V. L.; Cramer, S. P.; Yachandra, V. K., J. Am. Chem. Soc ., 2004, 126 , 9946.
62 Glatzel, P.; Bergmann, U., Coord. Chem. Rev. 2005, 249 , 65.
63 Carra, P.; Fabrizio, M.; Thole, B. T., Phys. Rev. Lett. 1995, 74 , 3700.
64 Reprinted from Co-ordination Chemistry Reviews, 249, (1-2), Glatzel, P., Bergmann, U., High resolution 1s core hole X-ray spectroscopy in 3d transition metal
50 Chapter 2 X-ray Techniques
complexes - electronic and structural information., 65. Copyright 2005, with permission from Elsevier
65 Reprinted with permission from Glatzel, P.; Singh, J.; Kvashnina, K. O.; van Bokhoven, J. A., J. Am. Chem. Soc., 2010, 132 , 2555. Copyright 2010, American Chemical Society.
66 Smolentsev, N.; Sikora, M.; Soldatov, A. V.; Kvashnina, K. O.; Glatzel, P., Phys. Rev. B, 84 , 235113.
67 Smolentsev, G.; Soldatov, A. V.; Feiters, M. C., Phys. Rev. B, 2007, 75 , 144106
68 Kas, J. J.; Rehr, J. J.; Soininen, J. A.; Glatzel, P., Phys. Rev. B, 2011, 83 , 235114.
69 Rehr, J. J.; Kas, J. J.; Vila, F. D.; Prange, M. P.; Jorissen, K., Phys. Chem. Chem. Phys., 2010, 12 , 5503.
70 van der Veen, R. M.; Kas, J. J.; Milne, C. J.; Pham, V.-T.; Nahhas, A. E.; Lima, F. A.; Vithanage, D. A.; Rehr, J. J.; Abela, R.; Chergui, M., Phys. Chem. Chem. Phys. 2010, 12 , 5551.
51
52 Chapter 3 Experimental
3 Experimental 3.1 Introduction This chapter will detail the synthesis of compounds that were synthesised for use in this thesis. In order to have a range of oxidation states, geometries and ligand types it was necessary to synthesise some known compounds in addition to purchasing more common materials commercially, such as the metal oxides. However due to experimental difficulties when at the synchrotron not all of the compounds which were synthesised were successfully measured, therefore these syntheses are not included.
The methods used to carry out the catalytic experiments are described here. Additionally information about the experimental techniques of the XANES and RIXS experiments is described here. Details of the FEFF9 calculations are also included in this chapter.
3.2 Instrumentation and general techniques Unless otherwise stated all syntheses and related characterisations were carried out in an inert atmosphere (using either nitrogen or argon) using Schlenk line techniques and an argon filled glove box. Chemicals were purchased from Sigma-Aldrich and Strem Chemicals.
NMR spectra were recorded on a Bruker AV-300 spectrometer in the solvents indicated at 298 K. Chemical shifts for proton and carbon spectra are reported on the δ scale in ppm and were referenced to the residual solvent. The coupling constants (J) are measured in Hertz (Hz). Infrared spectra of solid samples were obtained using a Thermo Nicolet 380 FT-IR spectrometer.
All solvents were dried by distillation under an inert atmosphere before use along with the following drying agents; toluene was distilled over sodium, diethyl ether over sodium and benzophenone, tetrahydrofuran (THF) over sodium and benzophenone, light petroleum ether (40-60) over sodium, benzophenone and diglyme and dichloromethane (DCM) over calcium hydride.
53 Chapter 3 Experimental
3.3 Synthesis of tungsten reference compounds
3.3.1 WOCl 4
WO 3 was synthesised from Na 2WO 4 and hydrochloric acid according to literature methods 1.
Thionyl chloride (60 ml) was added to the yellow WO3 (4.5g, 0.02 mol) directly by distillation and the orange solution was refluxed overnight. The remaining solvent was removed by distillation and the red-brown solid was dried under vacuum 2. Whilst this solid was used in further reactions for the RIXS experiments it was purified by sublimation at 80°C under vacuum conditions.
Melting point: 210°C (corresponds to literature)
3.3.2 WCp*Me 4 3 Firstly W(CO) 3MeCp* was prepared .
W(CO) 6 (6 g, 16.8 mmol) and MgClCp*.(C 4H8O) (4.56 g, 16.8 mmol), both white solids, were refluxed in dimethylformamide (DMF) for 1 hour to form a yellow solution. The DMF was removed and the yellow-orange residue was dissolved in tetrahydrofuran (THF) and refluxed for 2 hours with an excess of MeI (3.42 g, 24 mmol). The THF was removed by vacuum to give an oily yellow substance. The oil was repeatedly washed with petroleum ether (40-60°, 3 x 100 ml) and the resultant yellow solution was filtered. The solvent was removed from the solution by vacuum to give a yellow solid as the product (4 g, 9.6 mmol, yield = 57%).
1 W(CO) 3MeCp* : H (300 MHz, C 6D6): 1.45 (15H, s, C 5Me 5), 0.27 (3H, s, CH 3)
4 This product was then used to prepare [WCp*Cl 4] .
A solution of W(CO) 3MeCp* (4 g, 9.6 mmol) was dissolved in 20 ml dichloromethane
(DCM) and slowly added dropwise at room temperature to a stirred suspension of PCl 5 (4.98 g, 24 mmol) in DCM upon which CO was evolved and the solution turned orange then red. The solution was refluxed for 6 hours. The solution was then filtered and the red-orange solid retained. The solid orange-red product (1.3 g, 2.8 mmol) was washed with DCM (20 ml), THF (20 ml) and diethyl ether (20 ml) before being dried i n vacuo . The washings were retained and more red crystals later formed.
1 [WCp*Cl 4]: H (300 MHz, CD 2Cl 2): 13.7 (broad, C 5Me 5)
54 Chapter 3 Experimental
A 3M solution of MeMgBr (3.75 ml, 11 mmol) was diluted with 30 ml THF and stirred as 4 [WCp*Cl 4]2 (1.3 g, 2.8 mmol) was added as a solid . This red solution was left to stir for an hour after which it had become green. The solvent was removed by vacuum and the yellow-green residue was washed with petroleum ether (40-60°, 3 x 20 ml) to extract the WCp*Me 4. The solution was filtered and the solvent removed by vacuum to give a yellow-green powder. The solid was then recrystallised by slow cooling in the minimum volume petroleum-ether (40-60°).
1 H (300 MHz, CD 2Cl 2): 39.1 (broad peak, C 5Me 5). 5 1 Literature : H (CD 2Cl 2): 39.7 (broad)
WCl (N-4tBuC H ) 3.3.3 4 6 4
A modified literature method was followed for this reaction 5. The reaction was carried out under argon using dry solvents and reagents.
First, 4-tbutylphenyl isocyanate was synthesised as follows: 4-tbutylaniline (3g, 0.02 mol) was mixed with triethylamine (4g, 0.04 mol) in toluene (20 ml) and cooled to 0°C. To this orange mixture a solution of 20% phosgene in toluene (0.02 mol, 9.89g) was added slowly dropwise, forming a pale precipitate. The reaction mixture was allowed to warm to room temperature and stirred for one hour. It was then heated to 80°C for 1 hour before being left to cool. The mixture was then filtered through celite resulting in a yellow solution. The precipitate and celite was washed with light petroleum ether (40-60). The solution was transferred to a distillation apparatus. The solvents were removed by vacuum.
4-tbutylphenyl isocyanate: 1 H NMR (300 MHz, CDCl 3): 7.25 (2H, d, Ar), 6.94 (2H, d, Ar), 1.23 (9H, s, C(Me) 3)
t WOCl 4 (1.84 mmol) was stirred in octane (20 ml) to which freshly distilled 4- butylphenyl isocyanate (1.84 mmol) was added. This reaction mixture was refluxed overnight, turning brown from orange. After cooling the solvent was filtered off and the brown product was washed with toluene (10 ml) and then light petroleum ether (2 x 10 ml) before being dried under vacuum.
WCl (N-4tBuC H ) : 4 6 4 1 t H NMR (300 MHz, C 6D6): 7.16 (d, 2H, Ar), 6.91 (d, 2H, Ar), 1.11 (s, 9H, Bu)
55 Chapter 3 Experimental
3.4 Synthesis of rhenium reference compounds
3.4.1 [Re{(o-(HN)2C6H4)3][ReO 4] Tris(o-phenylenediamido)rhenium (VII) perrhenate was prepared according to the literature method 6 with minor modifications. The reaction was carried out under argon using tetrahydrofuran (THF) and diethyl ether distilled over sodium. The product is stable in air.
Re 2O7 (0.5 g, 1.03 mmol) was suspended in ~10 ml THF and stirred. To this was added a colourless solution of 1,2-phenylenediamine (0.75 g, 7 mmol) dissolved in THF which turned the reaction mixture dark green and this was stirred overnight. The solvent was then evaporated, the dark green solid washed with ether (3 x 5 ml) and dried under vacuum. The product was then extracted into acetone using a Soxhlet apparatus over 3 days to give a very dark green solution. The solvent was removed by vacuum to give a dark green solid (1.05 g, 1.9 mmol), [Re(o-(HN)2C6H4)3][ReO 4] in 92% yield. The product was used in the following reaction but was not used for RIXS due to the presence of 2 different rhenium species.
IR (KBr disc) / cm -1: 3272 [ ν(N-H)], 1552 [v(C=C)], 905 [v(Re=O)], 573 [v(Re-N)]
3.4.2 Re{(o-(HN)2C6H4)3
This reaction to form the paramagnetic Re(o-(HN)2C6H4)3 complex was carried out according to the literature method 6 with minor modifications. Sodium powder (0.27 g) in 10 ml THF was added to a stirred solution of [Re(o-
(HN)2C6H4)3 ][ReO 4] (1.95 g, 2.6 mmol) in 20 ml THF. After 24 hours the solvent was removed in vacuo to give a dark red/purple solid which was recrystallised in THF-light petroleum ether overnight. The solution was filtered off and retained for a second crop. The solid (0.75 g, 1.5 mmol, 58% yield) was dried under vacuum.
IR (KBr disc) / cm -1: 3270 [ ν(N-H)], 1554 [v(C=C)], 573 [v(Re-N)] Literature values 6 cm -1 : 3293, 1563, 573
3.5 Tungsten imido catalysis
WCl 6 (0.2 mmol) was added to an argon charged round bottom flask (RBF) followed by chlorobenzene (3 ml), aniline (0.2 mmol), triethylamine (0.3 mmol) with stirring. This was the starting solution. To form the end point 1-hexene (0.2 ml) and ethylaluminium dichloride (EADC) in toluene (2.2 mmol) was added, and left to stir under argon for ~1
56 Chapter 3 Experimental hour. The solution was filtered into an argon filled solution cell, at either the starting or end point. This was then used in the x-ray experiments.
Modifications: - In some experiments (described in Chapter 7) the aniline was replaced by 4- tbutylaniline. - This method is for the mono(imido) tungsten catalyst, for the bis(imido) two
equivalents of aniline were used per WCl 6.
3.6 RIXS experiments
The L 3 valence band RIXS spectra were collected at ID26 at the European Synchrotron
Radiation Facility (ESRF) between December 2009 and July 2011. The L β2 RIXS were collected at the SuperXAS beamline at the Swiss Light Source (SLS) in July 2010.
Figure 3.1 - Diagram of multi-crystal RIXS spectrometer set-up at ID26 7
3.6.1 Experimental set-up December 2009 (ESRF) The ring current was 60 to 90 mA with a filling mode of 16 bunch. The beamsize was 1mm x 0.5 mm on the sample. Pd/Cr 2.5 mrad mirrors were used to suppress higher
57 Chapter 3 Experimental harmonics. 600 µm attenuators were used. A Si (311) monochromator was used to select the incident energy.
An x-ray emission spectrometer using 3 Si (111) crystals at (555) reflections was used to select the emission energies with an Avalanche Photodiode detector. The overall energy resolution was 1.1 eV.
The elastic peak was used to determine the absolute energy in the energy transfer direction.
3.6.2 Experimental set-up December 2010 (ESRF) The ring current was 170-200 mA with a filling mode of 7/8 +1. The beamsize was 0.2mm x 0.5 mm on the sample. Pd/Cr 2.5 mrad mirrors were used to suppress higher harmonics. 600 µm attenuators were used. A Si (311) monochromator was used to select the incident energy.
An x-ray emission spectrometer using 5 Si (111) crystals at (555) reflections was used to select the emission energies with an Avalanche Photodiode detector.. The overall energy resolution was 1.7 eV.
The elastic peak was used to determine the absolute energy in the energy transfer direction.
3.6.3 Experimental set-up July 2011 (ESRF) The ring current was ~170 mA with a filling mode of 7/8 multibunch. The beamsize was 0.2 mm x 0.5 mm on the sample. Pd/Cr 2.5 mrad mirrors were used to suppress higher harmonics. 600 µm attenuators were used. A SI (311) monochromator was used to select the incident energy.
An x-ray emission spectrometer using 5 Si (111) crystals at (555) reflections was used to select the emission energies with an Avalanche Photodiode detector. The overall energy resolution was 1.5 eV.
The elastic peak was used to determine the absolute energy in the energy transfer direction.
58 Chapter 3 Experimental
3.6.4 Experimental set-up July 2010 (SLS) A Si(111) double crystal monochromator was used to select the incident energy and Rh coated mirrors were used to suppress higher harmonics. An emission spectrometer using 5 Si(111) crystals was used to select the emission energy. The overall energy resolution was 1.7 eV.
3.6.5 Sample preparation - powders Powder samples were made into 10 mm diameter thin (~0.5 mm) pellets and sealed onto a stainless steel sample plate with kapton tape in a nitrogen filled glove box. They were combined with boron nitride which had previously been heated under vacuum and then stored under nitrogen. They were then stored in a nitrogen filled vessel until mounting onto the cryostat arm and were then immediately placed inside the cryostat which was then evacuated and part filled with helium before being cooled to ~5 K.
The powder samples were cooled cryogenically to ~5 K using liquid helium. The oxides were measured outside of the cryostat due to practical constraints. Beam damage tests were carried out before each RIXS plane and the sample plate was moved every 5-20 minutes depending on beam damage test results so the beam was on fresh areas of the sample each time.
3.6.6 Sample preparation - solutions As described in 3.5 the catalytic solutions were prepared in the chemistry laboratory at the ESRF using an argon filled glove box, schlenk techniques (under argon), and anhydrous solvents from Sigma-Aldrich. All glassware was oven dried overnight at 200°C.
Solutions were measured in a solution cell with a circular window with a diameter of 2.5 cm and an internal volume of approximately 3 ml. The optimised measurements to prevent damage consisted of the sample stage motors moving constantly around the window and VB RIXS scans lasting approximately 2 hours.
3.7 FEFF9 calculations 3.7.1 Overview FEFF is a program designed for carrying out ab initio multiple scattering calculations of x-ray absorption and x-ray emission spectra for clusters of atoms, in addition to various other spectra. For this project a new extension of FEFF9, which includes both
59 Chapter 3 Experimental intermediate and final state core hole effects, was used to simulate the RIXS data which is not currently available in the general FEFF9 release.
There were a number of problems running the FEFF code for the materials in this thesis, so whereas FEFF is usually intended as a ‘parameter-free’ method, for this work each calculation needed a significant amount of individual attention. For example one common problem with FEFF is that it does not always accurately calculate the Fermi level within 2-3 eV, but with many of the tungsten and rhenium compounds it was often out by 10-15 eV. In addition the default core-hole values in FEFF did not always give optimal results, so literature values were substituted into code for some of the molybdenum RIXS calculations. There were also full potential issues which limited the success of the calculations. In addition there are obviously factors which FEFF does not account for, such as multiplet effects, so there will often be some differences between theory and experiment, even when optimised.
3.7.2 FEFF parameters for Mo L α RIXS calculations The co-ordinates of the crystal structure for each input structure. The main parameters used are summarised below. SCF refers to the self consistent field radius, centred on the absorbing atom, in angstroms. FMS refers to the full multiple scattering radius, centred on the absorbing atom, in angstroms. In order to run a FEFF calculation an input file must be created containing the atomic co-ordinates and a number of different cards which control key parameters. A selection of the most important cards used in a RIXS calculation is shown below in Table 3.1.
Table 3.1 - Key variables in FEFF input file for RIXS calculations Card Main usage EDGE Set the absorption and emission edges EXCHANGE Set the potential models, edge shift and broadening POTENTIALS Set the number of unique potentials, add stoichiometry, limit the angular momentum SCF Set the self-consistent field radius for the calculation FMS Set the full multiple scattering radius for the calculation COREHOLE Choose presence of a corehole and add screening RIXS Set broadening in incident energy and energy transfer directions, add energy shift EGRID Customise the RIXS energy grid, include different energy steps for different energy ranges
60 Chapter 3 Experimental
XES Set energy range and step size for XES calculation XANES Set energy range and step size for XANES calculation. The XANES output will then be convoluted with the XES calculation in the RIXS module
8 The M4 and M 5 corehole broadenings were taken from the literature , as they gave better results than the pre-set values in the FEFF code.
The stoichiometry was set in the POTENTIALS card, as well as the angular momentum for each atom in the SCF and FMS modules. Both dipole and quadrupole transitions were included.
MoO3: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr):0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 8 Å - FMS: 10 Å - Unique potentials: 3; absorbing tungsten, non-absorbing tungsten, oxygen - RIXS broadening: 0.1, 0.1 - Corehole: Included, RPA screened
MoO 2: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr):0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 8 Å - FMS: 10 Å - Unique potentials: 3; absorbing tungsten, non-absorbing tungsten, oxygen - RIXS broadening: 0.1, 0.1 - Corehole: Included, RPA screened
Na 2MoO 4: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
61 Chapter 3 Experimental
- E0 correction (Vr):0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 6 Å - FMS: 8 Å - Unique potentials: 3; absorbing tungsten, non-absorbing tungsten, oxygen - RIXS broadening: 0.1, 0.1 - Corehole: Included, RPA screened
MoS 2: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr):0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 8 Å - FMS: 10 Å - Unique potentials: 3; absorbing tungsten, non-absorbing tungsten, oxygen - RIXS broadening: 0.1, 0.1 - Corehole: Included, RPA screened
3.7.3 FEFF parameters for W and Re L 3 valence band RIXS calculations The co-ordinates of the crystal structure for each input structure. The main parameters used are summarised below. SCF refers to the self consistent field radius, centred on the absorbing atom, in angstroms. FMS refers to the full multiple scattering radius, centred on the absorbing atom, in angstroms.
Other cards such as UNFREEZEF (unfreezes f-electron density) and FOLP (changes the factor by which the muffin-tin radii are overlapped) were sometimes used to improve results. The stoichiometry was set in the POTENTIALS card, as well as the angular momentum for each atom in the SCF and FMS modules.
WO 3: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 9.0 - Pure imaginary potential (Vi): -2.37
62 Chapter 3 Experimental
- Background function potential model (ixc0): 2 - SCF: 8.4 Å - FMS: 10.4 Å - Unique potentials: 3; absorbing tungsten, non-absorbing tungsten, oxygen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
WO 2: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 0.5 - Pure imaginary potential (Vi): -2.5 - Background function potential model (ixc0): 2 - SCF: 8.0 Å - FMS: 10.0 Å - Unique potentials: 3; absorbing tungsten, non-absorbing tungsten, oxygen - RIXS broadening: 0.02 0.02 - Corehole: Included, RPA screened
Li 2WO 4: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 2.5 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 6.0 Å - FMS: 8.0 Å - Unique potentials: 4; absorbing tungsten, non-absorbing tungsten, oxygen, lithium - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened - There were problems calculating this structure, the XES calculated by FEFF gave negative intensity in parts, so the LDOS for the absorbing W d orbitals was used to replace this in the FEFF calculation
WOCl 4: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 2.5 - Pure imaginary potential (Vi): -2
63 Chapter 3 Experimental
- Ground state potential (ixc0): 2 - SCF: 6.0 Å - FMS: 8.0 Å - Unique potentials: 4; absorbing tungsten, non-absorbing tungsten, oxygen, chlorine - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
WO 2Cl 2(dme): - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 2 - Pure imaginary potential (Vi): -2.5 - Background function potential model (ixc0): 2 - SCF: 6.0 Å - FMS: 8.0 Å - Unique potentials: 6; absorbing tungsten, non-absorbing tungsten, oxide ligand, oxygen from dme ligand, carbon, hydrogen - RIXS broadening: 0.02, 0.02, plus -5 edge shift - Corehole: Included, RPA screened
t t W(NH Bu) 2(N Bu) 2: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 8.0 Å - FMS: 10.0 Å - Unique potentials: 6, absorbing tungsten, non-absorbing tungsten, nitrogen (imido), nitrogen (amido), carbon, hydrogen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
WCp*Me 4: - Potential model (ixc): 5 (Partially nonlocal: Dirac–Fock for core + HL for valence electrons + a constant imaginary part)
- E0 correction (Vr): 0.3
64 Chapter 3 Experimental
- Pure imaginary potential (Vi): -2.5 - Background function potential model (ixc0): 5 - SCF: 5.0 Å - FMS: 5.74 Å - Unique potentials: 15 (maximum), absorbing tungsten, hydrogen and 13 different carbon environments - RIXS broadening: 0.1, 0.1 - Corehole: Included, RPA screened
ReO 3: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 2.5 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 6.0 Å - FMS: 9.0 Å - Unique potentials: 3; absorbing rhenium, non-absorbing rhenium, oxygen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
ReO 2: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 0.5 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 7.0 Å - FMS: 9.0 Å - Unique potentials: 3; absorbing rhenium, non-absorbing rhenium, oxygen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
ReCl 3(PPh 2Me)3: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2
65 Chapter 3 Experimental
- SCF: 5.0 Å - FMS: 6.0 Å - Unique potentials: 5; absorbing rhenium, chlorine, phosphorus, carbon, hydrogen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
ReCl 3O(PPh 3)2: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 0 - Pure imaginary potential (Vi): -1.8 - Background function potential model (ixc0): 2 - SCF: 6.0 Å - FMS: 7.5 Å - Unique potentials: 7; absorbing rhenium, non-absorbing rhenium, chlorine, phosphorus, oxygen, carbon, hydrogen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
Re(o-pda) 3: - Exchange model (ixc): 2 (ground state + a constant imaginary part)
- E0 correction (Vr): 0 - Pure imaginary potential (Vi): -2 - Background function potential model (ixc0): 2 - SCF: 6.0 Å - FMS: 7.0 Å - Unique potentials: 4; absorbing rhenium, nitrogen, carbon, oxygen - RIXS broadening: 0.02, 0.02 - Corehole: Included, RPA screened
66 Chapter 3 Experimental
3.8 References
1 Brauer, G., Handbook of Preparative Inorganic Chemistry Vol II 2 nd Edition , 1965, Academic Press Inc
2 Colton, R., Tomkins, I. B., Aust. J. Chem , 1965, 18, 447
3 Mahmoud, K. A., Rest, A. J., Alt, H. G., Eichner, M. E., Jansen, B. M., J. Chem. Soc. Dalton Trans. 1984 , 175
4 Liu, A. H., Murray, R. C., Dewan, J. C., Santarsiero, B. D., Schrock, R. R., J. Am. Chem. Soc., 1987, 109, 4282
5 Schrock, R. R., DePue, R. T., Feldman, J., Yap, K. B., Yang, D. C., Davis, W. M., Park, L., DiMare, M., Schofield, M., Organometallics, 1990 , 9, 2262
6 Danopoulos, A. A., Wong, A. C. C., Wilkinson, G., Hursthouse, M. B., Hussain, B., J. Chem. Soc., Dalton Trans. , 1990 , 1, 315
7 Glatzel, P.; Sikora, M.; Smolentsev, G.; Fernández-García, M., Catal. Today 2009, 145 (3-4), 294-299.
8 Mårtensson, N.; Nyholm, R., Phys. Rev. B,, 1981, 24, 7121.
67
68
Chapter 4 Molybdenum Reference Studies
4 Molybdenum Reference Studies 4.1 Abstract A series of molybdenum reference compounds were studied using Resonant Inelastic X- ray Scattering (RIXS). The 2p 3/2 3d Mo RIXS planes show a rich structure with considerable more spectral information than in conventional X-ray Absorption Near Edge Structure (XANES) spectroscopy. The spectra were simulated using FEFF9 giving generally good agreement and detailed electronic information can be derived. The reference materials serve as a starting point for detailed electronic and geometric investigations of a broad range of compounds. Particularly, this can provide insights in the properties and performance of unknown and changing materials like those in catalysis.
4.2 Introduction Resonant Inelastic X-ray Scattering (RIXS) is a powerful tool in studying the electronic density of states whilst offering high spectral resolution 1,2. It directly probes the unoccupied and occupied density of states, usually studied by X-ray Absorption Spectroscopy (XAS) and X-ray Emission Spectroscopy (XES) respectively. The emitted photon energies are measured as a function of the absorption energies, in the X-ray absorption near edge (XANES) region, identifying their interdependencies and thereby uncovering additional features. It has already been applied successfully for the 3d transition metal complexes 3, where it has been used to reveal much greater detail in the K edge spectral shape 4 and thereby revealing information on, for example, the metal oxidation state, the local symmetry and crystal field splittings. 6 There are however few studies on 4d or 5d metal systems 1,5.
In a RIXS experiment a secondary spectrometer with an energy bandwidth similar to the lifetime broadening is used to selectively measure the emission energy 6. This allows the superior resolution seen in RIXS compared to normal XANES, since the broadening of the spectra is governed by the final state lifetime, rather than the initial state lifetime. For the Mo L RIXS as used in this work, emission spectra are recorded at incident energies along the L 3 edge region. The transitions involved are shown in Figure 4.1 as an energy level diagram. We are looking at the absorption of a photon, of energy , which promotes an n+1 electron from a 2p orbital to the 4d shell, creating a 2p 3/2 4d intermediate state. An electron from the 3d shell then drops down to fill the core hole, emitting a photon of energy
69
Chapter 4 Molybdenum Reference Studies
as it does so, and leaving a 3d hole in the final state. The spectra cover the M 4 and M 5 n+1 n+1 emission edges, so the 3d 3/2 4d and 3d 5/2 4d final states respectively. The difference between this final state and the initial state is the overall energy transfer of the system. The data is generally presented as a 2D contour plot, with the energy transfer plotted against the incident energy so that both are relative to the ground state.
Ground Intermediate Final State State State
5 10 n+1 2p 3d 3/2,5/2 4d Energy 3d 3/2,5/2 2p 3/2 Decay 2p 3/2 4d Absorption hν=
hν= Ω 6 9 n+1 2p 3d 3/2,5/2 4d
Energy Transfer Ω ω 6 10 n - 2p 3d 3/2,5/2 4d
Figure 4.1 - RIXS energy scheme for 2p 3/2 3d 3/2,5/2 RIXS. The vertical axis represents the energy of the electron configuration. For simplicity only the shells which are changing in configuration are shown.
There have been an increasing number of studies on 3d transition metal RIXS 7,8 but as yet there are no examples of 4d transition metal RIXS in the literature, mainly due to the relative inaccessibility of the energies involved. The L edges of 4d transition metals lie between 2000 and 4000 eV: These soft X-rays require the use of vacuum or reduced/change atmosphere (e.g. He) sample chambers in order to carry out experiments, since only a few centimeters of air absorb most of the photons. The advantage with studying the L edge as compared to K edge is that we are directly probing the 4d valence shell. Our spectra will be much sharper than a normal L-edge XANES spectrum due to the reduced lifetime broadening of a 3d core hole compared to a 2p core hole. The lifetime of the intermediate 2p 3/2 core hole broadens the spectrum in the incident energy axis only. Broadening of the spectra in the emission or energy transfer direction is due to the final state lifetime, i.e. the lifetime of the 3d hole, which is much shorter and only 0.16 eV for a
70
Chapter 4 Molybdenum Reference Studies
3d 3/2 and 0.14 eV for a 3d 5/2 core hole compared to 1.74 eV for a 2p 3/2 hole. We may also see different features in the RIXS compared to the XAFS as this is a two-step process compared to a one step p to d transition for L-edge XANES, so different symmetry operations are active.
There are direct L-edge molybdenum XANES studies in the literature, particularly for the metal oxides 9,10 , and these provide an interesting comparison to look for any additional effects in the RIXS. The direct L-edge XANES spectra all show a characteristic “white line” due to the dipole allowed 2p to 4d transition, and the splitting of this peak can be directly related back to the geometry (for simple geometries) 11,12,13 .
Figure 4.2 - Normalised fluorescence yield XANES taken from Hu and Wachs 14 . Reprinted with permission from J. Phys. Chem., 1995 , 99, ( 27 ), 10901. Copyright 1995 American Chemical Society.
14 Figure 4.2 shows examples of molybdenum L 3 XANES spectra and it can be seen that for the sodium molybdate, with the central Mo atom in a tetrahedral environment, the intensity ratio is approximately 2 : 3, whereas for the molybdenum (VI) oxide (octahedral) the intensity ratio is reversed and the first peak is of greater intensity. This is consistent with similar studies in the literature. This peak results from the 2p to 4d transition so the
71
Chapter 4 Molybdenum Reference Studies splitting relates directly to the crystal field splitting and the intensity ratios of the two peaks give a measure for the orbital distribution and thus geometry (Figure 4.3).
dz2 ,dx2-y2 Energy
oct Octahedral
dxy , d xz , d yz
Energy dxy , d xz , d yz
tet Tetrahedral
dz2 , d x2-y2
Figure 4. 3 – Energy level diagram showing d-orbital crystal field splitting for octahedral and tetrahedral environments
The charge transfer multiplet theory 15,16 has been successful in modeling the L-edge XAS of mainly 3d transition metals. This theory models the metal complex using crystal field theory and takes into account the multiplet effects present, i.e. the interactions of the core hole and the valence electrons. Moreover, charge transfer and electron transfer correlations can be accounted for, when necessary. The Mo L-edge XANES spectra of the
4d transition metal Mo systems Na 2MoO 4 and (NH 4)2MoS 4 have been simulated well using 9 the multiplet approach, despite these being very covalent systems . An extensive L 2 L 3 edge XAS study on octahedral 3d and 4d metals 17 , including Mo, has demonstrated that for 4d transition metals the 2p spin-orbit coupling is large and an L 3:L 2 intensity ratio of ~2:1 is observed. It is shown that pd and dd multiplet effects and 4d spin-orbit coupling are causing the difference between L 3 and L 2, with the L 2 being the least affected by multiplet effects and thus most suitable for single particle simulation techniques (like the FEFF technique as used here, vide infra ). However, since the multiplet technique reduces the material to a point-group symmetry only, no full structures can be calculated and no information on orbital hybridization and ligand interaction is obtained. Since our main goal is to develop RIXS as a tool in (organometallic) catalysis, we aim to obtain insights in the origin of the different features observed. This understanding, including their dependence on structure and electronic properties, will help us to interpret the RIXS planes of unknown and changing structures, thereby providing property-performance relationships on working
72
Chapter 4 Molybdenum Reference Studies catalysts. Therefore, the FEFF9 18 approach, which does allow the calculation of full structural models, is the preferred theoretical method to simulate the experimentally obtained Mo L 3 edge RIXS spectra. The effect of multiplets for the RIXS experiments as performed here is checked using the multiplet approach 15,16 and thereby the validity of the use of FEFF evaluated.
We present the theoretical spectra calculated using the real space Green’s function (RSGF) code FEFF9 18 for a range of molybdenum L RIXS spectra. The specific code used in this work includes a new theoretical treatment of RIXS based on a real-space multiple-scattering Green’s function formalism, and includes the effects of both intermediate and final state core-hole interactions. Additional many-body effects can be included via a convolution of the single particle spectrum with an effective spectral function, however in the present article, we have neglected these effects. This new method allows us to simulate these spectra successfully for the first time using a convolution of an effective XAS spectrum and the XES spectrum of the complexes simply based on their crystal structures. Detailed electronic information, i.e. d orbital occupancy and energy splitting, has been derived for these reference materials and the knowledge obtained on the interpretation of the RIXS data will be extremely useful when going to less defined materials. The number of peaks present, their intensity ratios, and the splitting between them can reveal structural and electronic information for unknown compounds, which is desirable for changing catalyst structures, where this information cannot be obtained otherwise.
4.3 Experimental Molybdenum (IV) oxide, molybdenum (VI) oxide, molybdenum (IV) sulfide and sodium molybdate were purchased from Sigma Aldrich.
The 2p3d RIXS spectra were recorded at ID26 of the European Synchrotron Radiation
Facility (ESRF), Grenoble, France across the molybdenum L 3 edge. A cryogenically cooled fixed-exit beamline monochromator consisting of a pair of Si crystals in the (111) reflection was used. The incoming flux was 10 13 photons / s.
A complete in-vacuum curved-crystal X-ray emission spectrometer 19 in Johansson geometry was used in collecting these spectra. The first order reflection of a Si(111)
73
Chapter 4 Molybdenum Reference Studies crystal was employed in the spectrometer. The sample was positioned 27 cm in front of the diffraction crystal in order to obtain energy resolution. A CCD camera was used to collect the spectra. The full RIXS plane consists of 69 consecutive 2p3d RIXS spectra recorded over an incident energy range of 2518 to 2535 eV with a step size of 0.25 eV. The acquisition time for a single RIXS spectrum was 60 s. The maximum count rate was 600 counts/s and occurred at the peak of the Mo L α1 emission line, excited at 2525 eV (first resonance).
The energy calibration of the spectrometer emission energy was done relative to the L line measured on a Mo foil, using the absolute nominal crystal 2d lattice spacing and crystal detector distance. For the excitation energy axis the nominal values given by the monochromator were considered. In meantime, for consecutive measurements, we have learned that it is better to use the elastic scattering peak for a more accurate calibration (spectrometer and monochromator). Between the two approaches an energy shift of up ~1 eV was observed. Although the possible shift is of course systematic and the same for all samples studies here, it has to be taken into account when comparing these results to other experiments.
FEFF 9 was used to model the experimental RIXS spectra and to determine the density of states. FEFF calculations were carried out using a self-consistent field approximation of 8 Å and a full multiple scattering radius of 10 Å around the central Mo atom. Both dipole and quadrupole contributions were included in the calculations. A ground state potential model and a RPA screened corehole were used. The EXCHANGE card was used to reduce the spectral broadening to match the experiment. Crystal structures from the literature were used. Core-hole broadenings for the M 4 and M 5 edges were taken from the 20 literature . A ratio of 9:1 was used for the L 3 edge spin orbit coupling, i.e. 2p 3/2 3d 5/2 = 9,
2p 3/2 3d 3/2 = 1, as expected based on transition probabilities. In case different Mo sites are present in the structure, the simulated data represent an average of all. Multiplet calculations were performed using the CTM4XAS program, with parameters as indicated in the text. 21
4.4 Results
Figure 4.4 shows the experimental L RIXS spectrum for Na 2MoO 4 as a contour plot alongside the theoretical plot for the same compound as calculated by FEFF9.
74
Chapter 4 Molybdenum Reference Studies
Figure 4. 4 - Experimental L 1 RIXS plot (left) for Na 2MoO 4 with theoretical RIXS plot calculated using FEFF9 (right) for the same compound.
The main peak is split in 2 along the diagonal, with maxima at 2525 eV incident energy and 232 eV energy transfer and at 2527 eV incident energy and 233.5 eV energy transfer. This corresponds to the ligand field splitting of the 4d orbitals. The structure is approximately repeated along a 2 nd diagonal shifted with respective to the first by ~3eV towards larger energy transfer. The two diagonals are due to the 3d spin-orbit coupling where the stronger spectral features (at lower energy transfer energy) arise from the 3d 5/2 final state. Theoretically the intensity ratio of the 3d spin orbit coupling is calculated as 9:1, with the second diagonal barely visible (see theoretical picture). However, in the experiment the transitions to the 3d 3/2 are much more pronounced.
The main peaks along the diagonal can be compared to the direct L 3 edge spectrum as shown in Figure 4.2. In the XANES spectrum the intensity is higher for the lower energy peak; this is evident in the RIXS plots also, when taking into account both the area of the peak and the intensity. In the RIXS plot the splitting is complete, in contrast to the overlapping peaks in the normal L edge experiment, and their positions and intensities can thus be better identified.
75
Chapter 4 Molybdenum Reference Studies
Along the energy transfer axis we see the spin orbit coupling of the 3d band, the lower energy transfer peaks are due to 3d 5/2 emission, the higher energy transfer peaks are due to emission from 3d 3/2 emission; both are due to 2p 3/2 to 4d absorption. So we can assign the peak at 2525 eV incident energy and 232 eV energy transfer to the 2p 3/2 to 4d; 3d 5/2 to
2p 3/2 transitions. The peak at 2525 eV incident energy and 234.8 eV energy transfer is again caused by 2p 3/2 to 4d absorption, but with 3d 3/2 to 2p 3/2 emission. The energy transfer difference between these two features is approximately 2.8 eV which corresponds to the spin-orbit coupling of the 3d band, generally found around 3.2 eV 22 . The peak at 2527 eV incident energy and 233.5 eV energy transfer is again caused by the 2p 3/2 to 4d; 3d 5/2 to
2p 3/2 transitions, but it is at higher energies due to the splitting of the 4d band. We see another peak at the same incident energy but at ~3 eV higher in energy transfer, again due to 3d spin orbit coupling.
The RIXS is simulated well using the FEFF9.0 approach. The peak intensity ratios are different compared to the experiment. The density of states as calculated and from which the RIXS plane is simulated is shown in Figure 4.5. A degree of hybridization can be seen between the unoccupied molybdenum d-orbitals and the unoccupied oxygen p-orbitals, as well as a small Mo p contribution.
Figure 4.5 – Density of States plot for Na 2MoO 4 as calculated by FEFF9
The potential effect of multiplets on the 2p3d Mo RIXS was checked using the multiplet theory. Previous L edge work suggests that Mo 4+ and Mo 6+ have 50% reduced Slater integrals, 17 therefore, the 2p3d RIXS planes were calculated with (Slater integrals to 50%
76
Chapter 4 Molybdenum Reference Studies atomic) and without multiplets. The main effects seen in the RIXS planes are the 2p and 3d spin-orbit couplings, as observed in the experiment and reproduced with FEFF9. The multiplets effects appear only very small and essentially only split the main peaks (i.e. peaks along the diagonal) into two peaks. This is not visible in our experiment, with the current experimental broadening and the additional fact that the atomic Slater integrals are considerably reduced for high valent 4d systems. Hence, the FEFF9 code can be reliably used to calculate the RIXS planes as done here.
In Figure 4.6 we present the experimental and calculated L RIXS spectra for MoO 3 as 2D contour plots. The main peak occurs at 2525 eV incident energy and 232 eV energy transfer. The second peak consists of two features, both along the diagonal. One is at 2527 eV incident energy and 233.5 eV energy transfer, the other at 2529 eV incident energy and 235.5 eV energy transfer. These can both be attributed to the 2p 3/2 to 4d; 3d 5/2 to 2p 3/2 transitions, with 4d splitting. There are off diagonal peaks which are 3 eV higher in the energy transfer direction than the main peaks. These are due to 3d 3/2 to 2p 3/2 emission. We see extra splitting in the main peaks because the Mo atom is not in a perfect octahedral environment, so there is hybridisation of orbitals causing extra transitions. For lower symmetries there will be more splitting of the 4d band due to increased orbital hybridisation.
Figure 4. 6 - Experimental L RIXS plot (left) for MoO 3 with theoretical RIXS plot
calculated using FEFF (right) for the same compound.
77
Chapter 4 Molybdenum Reference Studies
The experimental and theoretical plots for molybdenum (VI) oxide again show generally good correlation in the number, relative position and splitting of peaks. For example the splitting of the peaks between 234 and 236 eV energy transfer is reproduced in the RIXS plot calculated by FEFF. The off-diagonal peaks are also both present and the positions as well as intensity ratios are in good agreement. The origin of all peaks can be derived using the density of states calculations, as shown in Figure 4.7. Significant hybridisation of the Mo d and O p unoccupied orbitals is also observed, for all features present.
Figure 4.7 - Density of States for MoO 3 as calculated by FEFF9
In Figure 4.8 the experimental and theoretical L RIXS spectra for MoO 2 as 2D contour plots are shown. MoO 2 has a distorted rutile bulk structure so the molybdenum atoms are in distorted octahedral geometry. It has an unusual feature for a molybdenum oxide; a short metal-metal bond (~2.5 Å) 23 . The main peak along the diagonal is split, due to 4d splitting, with the peak of one feature at 2523 eV incident energy and 229.5 eV energy transfer (2p 3/2 to 4d; 3d 5/2 to 2p 3/2 ) and the other at 2526 eV incident energy and 232 eV energy transfer (2p 3/2 to 4d; 3d 5/2 to 2p 3/2 ). There are low intensity off-diagonal peaks approximately 3-3.3 eV above main peaks in energy transfer, which corresponds to the 3d spin orbit coupling.
78
Chapter 4 Molybdenum Reference Studies
Figure 4. 8 - Experimental L RIXS plot (left) for MoO 2 with theoretical RIXS plot calculated
using FEFF (right) for the same compound.
There is good agreement between the experimental and theoretical plots in terms of number of and relative positions of peaks. The density of states is shown in Figure 4.9 and shows hybridization between the Mo d and O p unoccupied orbitals, for all orbitals probed.
Figure 4.9 - Density of States for MoO 2 as calculated by FEFF9
79
Chapter 4 Molybdenum Reference Studies
The central molybdenum atom in MoS 2 is surrounded by 6 sulfur atoms in a trigonal prismatic geometry 24 . The overall structure is hexagonal. Figure 4.10 shows the experimental L RIXS spectra for MoS 2 as a contour plot. The main peak is split into two features; one at 2522.5 eV incident energy and 229.5 energy transfer, and the other at
2524 eV incident energy and 230.8 eV energy transfer (both are 2p 3/2 to 4d; 3d 5/2 to 2p 3/2 transitions with 4d splitting). There are additional peaks approximately 3eV higher in the energy transfer direction, these are due to 3d spin orbit coupling. They are weaker than in the other spectra. This is generally well reproduced with the simulation shown alongside, including the 4d splitting as observed. The density of states is shown in Figure 4.11, which agrees well with DOS schemas reported in literature25 .
Figure 4. 10 - Experimental L RIXS plot (left) for MoS 2 with theoretical RIXS plot calculated using FEFF (right) for the same compound.
80
Chapter 4 Molybdenum Reference Studies
Figure 4.11 - Density of States plot for MoS 2 as calculated by FEFF9
4.5 Discussion The RIXS spectra as obtained in this study show the clear advantage over normal L edge XANES studies in their ability to completely resolve the features present, as well as revealing new features. The features at lower energy transfer diagonal, corresponding to a
3d 5/2 final state core-hole, are intense features. The higher energy transfer peaks, corresponding to a 3d 3/2 final state core-hole, are weak for all samples. This is mostly due to the different transition probability of M 4 compared to M 5 emission, as mentioned (9:1 theoretically). The intensity ratio however varies across our four compounds. For example 0 in the tetrahedral d Na 2MoO 4 we see more intense and better-resolved off-diagonal peaks than the other three compounds. This is mostly due to the counting statistics, where in tetrahedral compounds there are 6 electrons in the higher energy t 2g orbitals compared to
4 electrons in the lower energy e g, see Figure 4.2. For octahedral compounds this is exactly the reverse.
The intensity ratio between the two main white line features in the diagonal of the RIXS plane, as also seen in a direct L edge experiment, albeit less-resolved, is not the perfect and predicted 2:3 or 3:2 as based upon ideal tetrahedral and octahedral geometry (Figure 2). The main reason for this, as reproduced by the simulations and shown in the DOS plots, is the orbital hybridization with ligands, i.e. overlap of ligand orbitals with the metal 4d, thereby changing the density of states structure and intensity. For the tetrahedral
81
Chapter 4 Molybdenum Reference Studies complex the ligand O p orbital only hybridised with the first main feature, i.e. the first d orbital, and not with the second, whereas for the other (distorted) octahedral complexes, hybridisation with all Mo d orbitals was observed at different intensities per orbital.
In Table 4.1 some important features of the RIXS spectra are summarized. A clear trend can be seen with the position of the first peak, which is clearly at 2523 eV incident energy and 232 eV energy transfer for the Mo(IV) compounds, and 2535 eV and 230 eV for the Mo(VI) compounds. The RIXS, compared to the normal direct L edge measurement, allows a more reliable oxidation state determination since the features are now fully resolved.
IE and ET Formal 3d 4d position of Compound oxidation splitting Geometry splitting / first peak / state / eV eV eV
Na 2MoO 4 Mo (VI) 2525, 232 3.1 Tetrahedral 2.1 Distorted MoO 3 Mo(VI) 2525, 232 3.1 3.9 Octahedral Distorted Octahedral, MoO 2 Mo (IV) 2523, 230 2.8 3.1 short M-M bond Trigonal MoS 2 Mo(IV) 2523, 230 3.2 - prismatic
Table 4.1 - Summary of Mo RIXS spectra
The 3d splitting is the spin-orbit coupling and this is very similar for all complexes, it being the 3d spin-orbit coupling which is fully shielded in the 4d transition metal. The 4d splitting reflects the sample geometry and is sample and geometry dependent. The 4 d splitting for the tetrahedral complex is clearly lower than the distorted octahedral complexes, which is expected. The crystal field splitting for a tetrahedral complex will be 4 th ~ /9 of the size of the crystal field splitting for an octahedral complex, for atomic systems assuming perfect symmetry, but the distortion of the octahedral geometry here will affect
82
Chapter 4 Molybdenum Reference Studies the ratio, as well as ligand hybridization, influencing the electron density of the molecular orbitals.
4.6 Conclusions As can be seen from the experimental 2p3d RIXS plots, we can obtain greater spectral resolution compared to the L 3 fluorescence yield XANES and as a consequence can provide more accurate oxidation state determination from the position of the first peak. It also reveals additional off diagonal features that are not visible with XANES, giving insights in the geometry of the complex. The spectra can be simulated using FEFF9, with good agreement in number and relative positions of peaks, since the multiplets plays little to no role in these spectra, with their current resolution. These same techniques can now be applied to a wide range of materials in order to obtain detailed electronic and geometric information under in situ and operando conditions. For unknown structures and changing materials the RIXS planes can be used as a direct measure for accurate oxidation states, geometry and crystal field splitting, giving insights in possible ligands present. For the Mo L α RIXS specifically, due to the relatively low X-ray energies used, special experimental cells have to be developed to allow these measurements in reduced atmosphere chambers.
4.7 Notes This work has been submitted to the Journal of Physical Chemistry B for publication. The author list is as follows: Rowena Thomas †¤, Josh Kas ‡, Pieter Glatzel §, Frank. M. F. De Groot ⊥, Roberto Alonso
§ ¥ ¥ ¥ ‡ † Mori , Matjaž Kav i ,Matjaz Zitnik , Klemen Bucar , John J. Rehr , Moniek Tromp ¤*
† University of Southampton, Southampton, SO17 1BJ, United Kingdom ¤ Catalyst Characterisation, Chemistry Department, Catalysis Research Center, Technische Universität München, Lichtenbergstrasse 4, 85748 Garching, Germany, [email protected] ‡ Department of Physics, University of Washington, Seattle, Washington 98195, United States
83
Chapter 4 Molybdenum Reference Studies
§ European Synchrotron Radiation Facility (ESRF), BP 220, 38043, Grenoble, Cedex 9, France
LCLS, SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA ⊥ Inorganic Chemistry & Catalysis, Debye Institute for Nanomaterials Science, Utrecht University, Universiteitsweg 99, 3584 CG, The Netherlands ¥ J Stefan Institute, SI-1001, Ljubljana, Slovenia For this work the ESRF is gratefully acknowledged for provision of beamtime under long term project CH2681. MT and RJT acknowledge the EPSRC (Advanced Research Fellowship EP/E060404/1) for funding.
In this work the RIXS experiments were undertaken by Moniek Tromp and co-workers in 2008, but the I processed the data, and carried out the FEFF calculations and analysis.
4.8 References
1 Glatzel, P.; Sikora, M.; Smolentsev, G.; Fernández-García, M.; Catal. Today , 2009 , 145, (3-4), 294.
2 Kotani, A.; Shin, S., Rev. Mod. Phys., 2001 , 73 , 203.
3 de Groot, F. M. F.; Glatzel P.; Bergmann, U.; van Aken, P. A.; Barrea, R. A.; Klemme, S.; Hävecker, M.; Knop-Gericke, A.; Heijboer, W. M.; Weckhuysen, B. M., J. Phys. Chem. B, 2005, 109 , 20751.
4 Kotani, A.; Matsubara, M.; Uozumi, T.; Ghiringhelli, G.; Fracassi, F.; Dallera, C.; Tagliaferri, A.; Brookes, N. B.; Braicovich, L., Rad. Phys. Chem., 2006 , 75 , 1670.
5 Glatzel, P.; Singh, J.; Kvashnina, K. O.; van Bokhoven, J. A, J. Am. Chem. Soc. , 2010 , 132 , 2555.
6 Glatzel, P.; Bergmann, U., Coord. Chem. Rev., 2005, 249 , 65.
7 Glatzel, P.; Yano, J.; Bergmann, U.; Visser, H.; Robblee, J. H.; Gu, W.; de Groot, F. M. F.; Cramer, S. P.; Yachandra, V. K.; J. Phys. Chem. Solids, 2005 , 66 , 2163.
8 Meyer, D. A.; Zhang, X.; Bergmann, U.; Gaffney, K. J.; J. Chem, Phys ., 2010 , 132 , 134502.
84
Chapter 4 Molybdenum Reference Studies
9 George, S. J.; Drury, O. B.; Fu, J.; Friedrich, S.; Doonan, C. J.; George, G. N.; White, J. M.; Young, C. G.; Cramer, S. P., J. Inorg. Biochem., 2009 , 103 , 157.
10 Bare, S. R.; Mitchell, G. E.; Maj, J. J.; Vrieland, G. E.; Gland, J. L., J. Phys. Chem., 1993 , 97 , 6048.
11 Evans, J.; Mosselmans, J. F. W., J. Phys. Chem., 1991 , 95 , 9673.
12 Aritani, H.; Tanaka, T.; Funabiki, T.; Yoshida, S.; Eda, K.; Sotani, N.; Kudo, M.; Hasegawa, S., J. Phys. Chem., 1996 , 100 , 19495.
13 Hu, H.; Wachs, I. E.; Bare, S. R. J. Phys. Chem. 1995 , 99 , 10897
14 Hu, H.; Wachs, I. E., J. Phys. Chem., 1995 , 99, 10897
15 Ikeno, H.; de Groot, F. M. F.; Stavitski, E.; Tanaka, I., J. Phys.: Condens. Matter , 2009 , 21 , 104208.
16 de Groot, F., Coord. Chem. Rev ., 2005 , 249 , 31.
17 de Groot, F. M. F.; Hu, Z. W.; Lopez, M. F.; Kaindl, G.; Guillot, F.; Tronc, M., J. Chem. Phys., 1994 , 101 , 6570.
18 Kas, J. J.; Rehr, J. J.; Soininen, J. A.; Glatzel, P., Phys. Rev. B., 2011, 83 , 235114.
19 Kav i , M.; Budnar, M.; Mühleisen, A.; Gasser, F.; Žitnik, M.; Bu ar, K.; Bohinc, R., Rev. Sci. Insrum., 2012 , 83 , 033113.
20 Mårtensson, N.; Nyholm, R., Phys. Rev. B, 1981, 24 , 7121.
21 Stavitski, E., de Groot, F. M. F., Micron 2010 , 41 , 687.
22 Fuggle, J. C.; Mårtensson, N., J. Electron Spectrosc. Relat. Phenom. 1980, 21 , 275.
23 Brandt, B. G.; Skapski, A. C., Acta. Chem. Scand., 1967 , 21 , 661.
24 Wildervanck, J. C.; Jellinek, F., Z. Anorg. Allg. Chem., 1964, 328 , 309.
25 Raybaudyz, P., Hafnery, J., Kressey, G., Toulhoatz, H. J. Phys.: Condens. Matter , 1997 , 9, 11107.
85
86 Chapter 5 Tungsten Reference Compounds
5 Tungsten Reference Compounds 5.1 Introduction In this chapter the high energy resolution fluorescence detected (HERFD) X-ray Absorption Near Edge Structure (XANES) and valence band resonant inelastic x-ray scattering (VB RIXS) spectra for a series of tungsten reference complexes will be discussed. The data was collected at ID26 at the European Synchrotron Radiation Facility (ESRF) in France for the Valence Band studies, and the SuperXAS beamline at the Swiss Light Source (SLS) for the L beta 2 studies.
Comparatively few XANES and RIXS studies exist for 5d metals compared to their 3d counterparts, and again there are relatively few studies carried out on the early transition metals. The existing studies are mainly core to core RIXS, but in these studies we are probing valence to core RIXS. XANES has become a more widely used technique in recent years, although the necessity for a high energy x-ray source limits the availability compared to other techniques which can be carried out in-house at any University. However XANES is commonly used as a fingerprint technique to see if a particular species or oxidation state is present and is rarely explored further. In this chapter we aim to look in more detail at the electronic and geometric effects which govern the shape of these L-edge tungsten XANES spectra. By using the HERFD technique we can uncover features that were previously unresolved in normal XANES. In addition we have the Valence Band RIXS for the same compounds; this enables us to probe the region around the valence band in more detail.
The HERFD spectra were collected by recording a XANES spectrum at a specific emission energy, in this case 10202 eV, in order to reduce the life time broadening of the core hole. A secondary monochromator set at this energy was used to select the emission energy at which the absorption spectrum is collected. In addition to this there was a fluorescence detector set up behind the analyser crystals (i.e. the secondary monochromator) which gives us the more usual, fluorescence detected, XANES.
Theoretical spectra were calculated using the real space Green’s function (RSGF) code FEFF9 1, which uses a new theoretical treatment of RIXS based on a real-space multiple-scattering Green’s function formalism and a quasiboson model Hamiltonian to account for the single-particle spectrum and multi-electron excitations respectively. This new method uses a convolution of the XAS and XES calculations for the complexes based on their crystal structures to calculate the RIXS planes.
87 Chapter 5 Tungsten Reference Compounds
5.2 Tungsten (IV) oxide Tungsten (IV) oxide has a rutile type structure which consists of chains of distorted 2 octahedra with a monoclinic unit cell (P2 1/c) , see Figure 5.1. Conventionally, the tungsten is in a +4 oxidation state and therefore is assumed to have two 5d electrons. However electron counting is a formalism and will not necessarily reflect the true electron distribution within the structure.
The distortion of the octahedral geometry allows metal-metal bonding along the c axis 3.
The single t 2g orbitals which lie along the edges of the octahedra interact with the equivalent orbitals on neighbouring metal atoms and can form σ bonding interactions, therefore the t 2g d orbitals can be thought of as split.
Figure 5.1 - Structure of tungsten (IV) oxide viewed down the b axis, blue represents tungsten atoms, red oxygen atoms. Created using Mercury 2.4 and crystal structure from Bolzman et al, 1995 4
Shown below is the HERFD spectrum obtained at ID26 for tungsten (IV) oxide. As a comparison the normal XANES (i.e. total of all emission energies) for tungsten (IV) oxide is shown alongside the HERFD in Figure 5.2. These were collected during the same experiment. It is clear that there is much greater energy resolution in the HERFD as evidenced by the greater number of features visible and the sharpness of these features. The first peak is the elastic peak. The main peak is split in two, with
88 Chapter 5 Tungsten Reference Compounds greater intensity at the higher energy part. This peak probes the 5d orbital, it is the white line caused by the 2p to 5d transition 5. The splitting of the peak indicates that there are two distinct energies at which 2p electrons are promoted to 5d. There is relatively little in the literature 6 on the splitting of this white line, especially compared to the number of papers discussing the splitting of the molybdenum L 3 white line and it’s relation to structure and geometry (see Chapter 4). It is generally not seen for tungsten 5 L3 edge XANES in the literature, without first taking the second derivative for example , illustrating the power of HERFD to uncover features hidden by the 2p lifetime broadening of normal L 3 XANES.
The HERFD plot for tungsten (IV) oxide is also shown in Figure 5.3 alongside the XANES plot simulated by FEFF 9 with an experimental broadening of -1.5 eV. The first, sharp peak in the experimental HERFD spectrum is the elastic peak, which corresponds to the elastic peak shown in the experimental RIXS (Figure 5.5). It represents elastic transitions, i.e. no net energy transfer. It is visible in the HERFD and not the normal XANES because the HERFD is measured at a specific emission line which enables this to be seen. The simulation shows generally good agreement with the experimental spectrum, particularly with respect to the splitting of the main peak, both the relative size of the peaks and the separation of the peaks is reproduced.
HERFD
10180 10200 10220 10240 10260 10280 Normal XANES
10180 10200 10220 10240 10260 10280 Energy (eV)
Figure 5.2 - HERFD (top) and normal fluorescence yield XANES (bottom) spectra for tungsten (IV) oxide
89 Chapter 5 Tungsten Reference Compounds
1 .2 Experimental Calculate d 1 .0
0 .8
0 .6
0 .4
Normalised absorption 0 .2
0 .0
10180 10200 10220 10240 10260 10280 10300 Incident E nergy (eV)
Figure 5.3 – L3 HERFD XANES spectrum for WO 2 taken at the valence band emission energy (10202eV) alongside the spectrum simulated using FEFF9
Figure 5.4 shows the density of states plot as calculated by FEFF 9 for tungsten (IV) oxide. It shows the predicted occupancy of different atomic orbitals within the complex at different energies. The energies above 0 eV (the Fermi level) represent the unoccupied density of states, i.e. the states into which the electrons are excited in a XANES experiment.
90 Chapter 5 Tungsten Reference Compounds
Figure 5.4 - Density of States plot calculated by FEFF for tungsten (IV) oxide, calculated using FEFF9
There is significant filling of tungsten d orbitals below the Fermi level (0 eV), which we do expect due to the 5d2 configuration. We see particularly high mixing of the oxygen p orbitals and the tungsten d-orbitals between -8 and -12 eV, forming the conductance band.
In Figure 5.5 is the experimental Valence Band RIXS spectrum for tungsten (IV) oxide. The data is presented as a 2D contour plot with the energy transfer plotted against the incident energy, with the contours and colours representing the intensity, from green at the lowest intensities, to red at the highest relative intensity. The plot has been calibrated so the elastic peak is at zero energy transfer as the elastic peak represents transitions with no net energy gain or loss.
There is a small peak close to the elastic peak at 10208 eV incident energy and 3 eV energy transfer, the most intense peak at 10211 eV incident energy and 6 eV energy transfer. There is an off diagonal peak at 10210 eV incident energy and 10 eV energy transfer. It can be seen that there is no clear gap between the lowest energy transfer feature and the elastic peak, indicating a metallic type electronic structure.
91 Chapter 5 Tungsten Reference Compounds
Figure 5.5 - Experimental L3 Valence Band RIXS 2D contour plot for WO 2
Figure 5.6 - Theoretical L3 Valence Band RIXS 2D contour plot for WO 2 as calculated by FEFF9
92 Chapter 5 Tungsten Reference Compounds
Also shown is the calculated VB RIXS spectrum, Figure 5.6, produced using a FEFF9 calculation which shows the 3 main features that we see in the experimental plot. The relative size and positions of these features seems in good agreement. The relative intensities of the features also appear to be comparable to the experimental plot.
If we look at the unoccupied density of states (i.e. above 0 eV in Figure 5.4) we can correlate these with the diagonal cut through the RIXS plot (which corresponds to the HERFD or constant emission energy) and it appears that the first peak in the RIXS (10207.7 in experiment) is due to a transition into the first unoccupied W d state, at ~0.5 eV and the second peak along the main diagonal (10211.1 eV incident energy in experiment) is due to the third peak in the unoccupied DOS at ~3.5 eV. From these values it appears that the FEFF calculation is out by ~0.5 in the splitting along this main diagonal when comparing to the centres of the experimental peaks.
A vertical cut through the RIXS plane (constant incident energy) will correspond to the filled density of states. If we look at the filled states, we have peaks at approximately 3, 6 and 10 eV energy transfer. The RIXS plot is likely to appear more broadened than the density of states, which have all experimental broadening removed. But if we refer to the zero energy transfer position, which will be the same in the calculated RIXS (Figure 5.6) as the density of the states (Figure 5.4), we can see that the peak at -3 eV, which is almost entirely W d, with a low p contribution from both the W and O atoms, corresponds to the lowest energy transfer peak. The second peak in the energy transfer direction therefore appears to be due to emission from a mostly O p state, with some W p and d hybridisation (at -6 eV in Figure 5.4). The third peak in the energy transfer direction corresponds to emission from the region of high W d and O p occupancy between -9 and -10 in the density of states.
5.3 Tungsten (VI) oxide In this section we discuss tungsten (VI) oxide, a monoclinic7 solid with the tungsten in a distorted octahedral environment (see Figure 5.7). The tungsten is in a +6 formal oxidation state so d 0 configuration.
93 Chapter 5 Tungsten Reference Compounds
Figure 5.7 - Structure of tungsten (VI) oxide viewed down the c axis, blue represents tungsten atoms, red oxygen atoms. Created using Mercury 2.4 and crystal structure from Loopstra et al 7.
The L3 HERFD XANES plot collected at ID26 can be seen in Figure 5.8 along with the plot calculated by FEFF9 for this compound. A good agreement is seen with the splitting of the peak.
XANES (FEFF) 1.0 HERFD (Exp.)
0.8
0.6
0.4
Normalised Absorption 0.2
0.0
10190 10195 10200 10205 10210 10215 10220 10225 10230 Energy (eV)
Figure 5.8 - L3 HERFD XANES spectrum for WO 3 taken at the valence band emission energy (10202 eV) alongside the spectrum simulated using FEFF9
94 Chapter 5 Tungsten Reference Compounds
Figure 5.9 - Density of States plot as calculated by FEFF9 for tungsten (VI) oxide. Key refers to atomic orbitals which contribute towards the overall bonding.
In Figure 5.10 is shown the experimental L 3 VB RIXS plot for tungsten (VI) oxide, which shows a broad peak which appears to be split, although this is not fully resolved. The main centres of intensity are at 10211 eV incident energy and 8 eV energy transfer, and at 10214 eV incident energy and 12 eV energy transfer. There is a clear gap between the elastic peak (at 0 eV energy transfer) and the inelastic peak, unlike the VB RIXS plot for tungsten (IV) oxide (Figure 5.5) where the lowest energy transfer peak is not completely resolved from the elastic peak, indicating a metallic like electronic structure.
The FEFF VB RIXS calculation (Figure 5.11) shows poor agreement when looking at the splitting of the main peak, which is 2.8 eV in the energy transfer direction in the experimental plot, and more than double this in the FEFF calculation. In the FEFF RIXS plot the splitting almost fully resolves the main feature into two separate peaks. The unoccupied density of states is in good agreement with the density of states calculated in the work of Smolentsev et al 8, who used a full potential density functional theory (DFT) approach, so it is unclear why there is such a big difference when looking at the 2D RIXS plots, whereas for the DFT calculations there was fairly good agreement in the RIXS planes. It appears that the energy difference between the filled and empty DOS is incorrect, so the band gap is incorrect, and this has been translated
95 Chapter 5 Tungsten Reference Compounds into the RIXS. It seems that part of the problem originates with the way FEFF calculates the RIXS plane vs. the HERFD and DOS.
Figure 5.10 - Experimental L3 Valence Band RIXS 2D contour plot for WO 3
Figure 5.11 - Theoretical RIXS plot for WO 3 as calculated by FEFF9
96 Chapter 5 Tungsten Reference Compounds
5.4 Lithium tungstate
In lithium tungstate (Li 2WO 4) the tungsten is surrounded by 4 oxygen atoms in a 2- tetrahedral environment to form a WO 4 subunit. The formal oxidation state is 6+, giving a d 0 configuration. The structure is illustrated in Figure 5.12.
Figure 5.12 - Structure of lithium tungstate, blue atoms represent tungsten, red oxygen and grey lithium
Figure 5.13 - Density of states plot for Li 2WO 4 as calculated by FEFF9
97 Chapter 5 Tungsten Reference Compounds
By examining the density of states plot (Figure 5.13) we can see there is significant O p and W d hybridisation between 10 and 7.5 eV below the edge, W p and O p hybridisation in the feature at -6.5 eV and again O p and W d hybridisation in the first unoccupied peak (at 0 eV), although the p contribution is relatively low here.
Shown below are the experimental W VB RIXS plot and the theoretical plot, calculated using an extension of FEFF9, for Li 2WO 4. There was difficulty calculating this structure as the code is still in the developmental phase. FEFF calculates the RIXS by convoluting the XAS and XES spectra, but in this case there was a problem with the XES spectra, causing a large negative intensity at one point for unknown reasons. Therefore the relevant area of the d DOS was used in place of the XES in order to calculate the RIXS. The density of states (Figure 5.13) clearly does not show the double feature present in the RIXS (Figure 5.14), despite all broadening being removed from the calculation, therefore it is also not present in the calculated RIXS plot (Figure 5.15). Although the splitting of the feature is not seen there is fairly good agreement in the energy transfer position of the peak, and some asymmetry can be seen in the peak shape.
Figure 5.14 - Experimental L 3 Valence Band RIXS 2D contour plot for Li 2WO 4
98 Chapter 5 Tungsten Reference Compounds
Figure 5.15 - Theoretical L 3 Valence Band RIXS plot for Li 2WO 4 as calculated by an extension of FEFF9
If we compare the RIXS (see Figure 5.14 and Figure 5.15) with the density of states plot shown in Figure 5.13 we can see that the gap between the highest occupied W d orbitals (between -8 and -10 eV) and the lowest unoccupied W d-orbital at 0 eV corresponds to the energy transfer gap between the elastic peak and the inelastic peaks in Figure 5.14. There is a large amount of hybridisation between the oxygen p orbitals and the W d-orbitals in the occupied density of states (between -8 and -10 eV), which reflects the molecular orbitals from which emission occurs. For lithium tungstate the Fermi level should be in the gap. The diagonal cut through the RIXS (CEE) reflects the unoccupied density of states. The FEFF RIXS simulation (Figure 5.15) does not show the same diagonal splitting that the experimental RIXS plot does (Figure 5.14) but the position in energy transfer agrees.
5.5 Tungsten (VI) oxotetrachloride
When considering the WOCl 4 molecule it is square pyramidal in geometry, with a tetragonal space group 9. When considering the bulk structure we find that these molecules are oxygen bridged, so each tungsten atom is surrounded by 2 oxygen atoms and 4 chlorine atoms in a distorted octahedron. Chains of these octahedra run parallel to the c axis.
99 Chapter 5 Tungsten Reference Compounds
Figure 5.16 - Structure of tungsten (VI) oxotetrachloride viewed down the a axis, blue represents tungsten atoms, red represents oxygen atoms, and green represents chloride atoms. Created using Mercury 2.4 and crystal structure from Hess and Hartung, 1966 9.
In Figure 5.17 a comparison of the normal XANES and the HERFD, collected during the same experiment is shown, alongside the XANES spectrum calculated using FEFF9. The first peak in the HERFD is the elastic peak, as it was measured with the emission energy fixed at the valence band. The HERFD shows a shoulder on the peak that is not visible in the normal XANES. The XANES shows fairly good agreement, although the shoulder is not as pronounced as in the HERFD experiment.
The density of states plot (Figure 5.18) shows the hybridisation present in the occupied orbitals, with the most significant overlap between Cl p, O p and W d. The Cl p orbitals also contribute to the unoccupied density of states, at around 0 eV.
100 Chapter 5 Tungsten Reference Compounds
Figure 5.17 - L 3 VB HERFD, normal XANES and XANES calculated by FEFF for
WOCl 4
Figure 5.18 - Local density of states for WOCl 4 as calculated by FEFF9
101 Chapter 5 Tungsten Reference Compounds
The experimental W L 3 VB RIXS plot is shown in Figure 5.19 and the theoretical plot calculated using an extension of FEFF9 is shown in Figure 5.20. There is fairly good agreement in the energy transfer position and shape of the peak for this compound.
We can correlate the density of states features to the position of the VB RIXS peak in the energy transfer position. The energy transfer gap appears to correlate well with the highest significant filled W d states, between -4 and -5 eV and the lowest unoccupied d states at around 2-3 eV, which corresponds to the peak in the VB RIXS, which is centred at 7.2 eV energy transfer. It appears that the filled states have a large amount of both O p and Cl p character, in addition to a small W p contribution. The unoccupied states are mostly W d but have a small amount of hybridisation with Cl p orbitals.
Figure 5.19 - Experimental L 3 Valence Band RIXS 2D contour plot for WOCl 4
102 Chapter 5 Tungsten Reference Compounds
Figure 5.20 - Theoretical L 3 Valence Band RIXS plot for WOCl 4 as calculated by an extension of FEFF9
5.6 Tungsten hexacarbonyl Tungsten hexacarbonyl has an orthorhombic (pnma) structure, with the tungsten atoms in octahedral geometry, as shown in Figure 5.21. The tungsten has a formal oxidation state of zero.
Figure 5.21 - Structure of W(CO) 6, blue represents tungsten, grey carbon, red oxygen
This calculation was particularly problematic and it is thought that there are full potential issues. FEFF uses muffin tin approximations for its treatment of atomic potentials which is not always successful, indicating that a more complex approach to the potentials is needed. From looking at the comparison between experimental HERFD and XANES and the XANES calculated by FEFF (Figure 5.22) it can be seen
103 Chapter 5 Tungsten Reference Compounds that FEFF does not replicate the experimental spectrum well, as there is splitting of the main feature, and it is broader than the experimental spectra, even when the broadening was adjusted in the calculation. The density of states was calculated for
W(CO) 6 using FEFF9 and is shown in Figure 5.23.
Figure 5.22 - Experimental L 3 VB HERFD and normal XANES, with XANES as calculated by FEFF9
104 Chapter 5 Tungsten Reference Compounds
Figure 5.23 - Density of states plot for W(CO) 6 as calculated by FEFF9
The experimental L 3 VB RIXS plot is shown in Figure 5.24 and the FEFF calculation in Figure 5.25. Although the energy transfer position for both is around 10 eV there is poor agreement in the splitting of the peak. In the experimental plot it is along the diagonal whereas in the FEFF calculation it is in the energy transfer direction.
Figure 5.24 - Experimental L 3 Valence Band RIXS 2D contour plot for W(CO) 6
105 Chapter 5 Tungsten Reference Compounds
Figure 5.25 - Theoretical L 3 Valence Band RIXS 2D contour plot for W(CO) 6
5.7 Bis(tert-butylamido) bis(tert-butylimido) tungsten
The structure of the WVI , d 0, bis(tert-butylamido) bis(tert-butylimido) tungsten complex is of the Pbcn space group. The solid state structure measured in the literature showed disorder over two equal sites, although the imido and amido groups are well distinguished, so the structure is a weighted average of the two 10 .
The experimental HERFD and normal XANES (collected in same experiment) are shown in Figure 5.26 - HERFD XANES at the valence band, normal fluorescence detected XANES and XANES calculated by FEFF9 for W(NHtBu)2(NtBu)2 alongside
106 Chapter 5 Tungsten Reference Compounds the XANES spectrum calculated using FEFF9. The splitting in the FEFF XANES is slightly less (by ~0.5 eV) and the broadening of this split peak could not be well reproduced. The splitting should correspond to the splitting of the unoccupied 5d orbitals.
The density of states (Figure 5.27) show particularly strong hybridisation between the W d orbitals and the imido N orbitals in the occupied states, before – 5 eV, and above this, still in the occupied states we see increased occupancy of the amino N orbitals and hybridisation with the W d and imido N orbitals. In the unoccupied states there is a small degree of hybridisation of both types of N p orbitals and the W p orbitals with the W d orbitals.
Figure 5.26 - HERFD XANES at the valence band, normal fluorescence detected t t XANES and XANES calculated by FEFF9 for W(NH Bu) 2(N Bu) 2
107 Chapter 5 Tungsten Reference Compounds
t t Figure 5.27 - Density of states for W(NH Bu) 2(N Bu) 2 as calculated by FEFF9
Figure 5.28 - Experimental L 3 Valence Band RIXS 2D contour plot for t t W(NH Bu) 2(N Bu) 2
108 Chapter 5 Tungsten Reference Compounds
t t Figure 5.29 - Theoretical L 3 Valence Band RIXS 2D contour plot for W(NH Bu) 2(N Bu) 2
The experimental (Figure 5.28 - Experimental L3 Valence Band RIXS 2D contour plot for W(NHtBu)2(NtBu)2) and theoretical (Figure 5.29) L 3 VB RIXS plots are shown above, and it can be seen that the energy transfer position is fairly well reproduced, but the splitting of the RIXS feature is not seen in the theoretical plot.
If we correlate the density of states plot to the RIXS we can see the ET gap between the elastic peak (i.e. 0 ET) and the inelastic peaks (centred at 7.2 and 9.5 in the experimental plot) corresponds to the transition between the highest occupied W d- orbital at -5 eV and the split W d peak in the unoccupied density of states (between 2.5 and 5 eV). Therefore the transition between the highest occupied W d and lowest energy unoccupied W d orbitals is approximately between 7.5 and 10 eV, corresponding to the ET position. The density of states shows that there is a significant amount of hybridisation between the N p orbitals and the W d orbital at -5 eV.
The unoccupied W d peaks between 2.5 and 5 eV show only a small contribution from N and W p orbitals. However the splitting evident in the unoccupied density of states is not evident in the RIXS calculation, there is no diagonal splitting visible.
109 Chapter 5 Tungsten Reference Compounds
5.8 Pentamethylcyclopentadienyl tetramethyl tungsten The structure reported by Kohler et al (1995) 11 was used to calculate the FEFF V simulations for WCp*Me 4, a W complex. The molecule is in a tetragonal pyramidal 12 arrangement , shown below in Figure 5.30 - Structure of WCp*Me 4
H3C W CH3
H3C CH3
Figure 5.30 - Structure of WCp*Me 4
The HERFD and normal XANES experimental spectra are shown in Figure 5.31 alongside the XANES spectra calculated by FEFF. The FEFF calculated XANES has some clear differences from the experimental plots, the two main peaks in the HERFD spectrum being further split into two. The overall splitting between the two main peaks is also larger than the HERFD splitting.
110 Chapter 5 Tungsten Reference Compounds
Figure 5.31 - HERFD XANES at the valence band, normal fluorescence detected
XANES and XANES calculated by FEFF9 for WCp*Me 4
Figure 5.32 - Density of states plot for WCp*Me 4 as calculated by FEFF9
111 Chapter 5 Tungsten Reference Compounds
The density of states is shown in Figure 5.32. The carbon p orbitals were not show for simplicity, as there were 13 unique carbon potentials set in the FEFF calculation, each one giving a different result.
If we compare the experimental RIXS plot (Figure 5.33) with the theoretical RIXS plot (Figure 5.34) we can see that the positions and splitting of the peaks in the FEFF calculation do not correspond exactly to the experimental peaks. The first peak in the experimental plot is centred at 5.8 eV energy transfer, but in the calculation it is at 4.1 eV. The splitting of the peaks in the energy transfer direction, however, is well reproduced, at around 3.4 eV for both plots. The splitting in the incident energy direction is less, by 1 eV, for the FEFF calculation than the experiment. It is also clear that the relative intensity of the two peaks is not well reproduced.
If we compare the W d DOS in Figure 5.32 to the FEFF calculation in Figure 5.34 we can see that if the highest occupied W d state (between -4 and -2 eV) corresponds to zero energy transfer (i.e. the elastic peak in Figure 5.33), therefore the two peaks will correspond to excitations into the unoccupied d states at 0 to 2 eV and at around 5 eV.
Figure 5.33 - Experimental L 3 Valence Band RIXS 2D contour plot for WCp*Me 4
112 Chapter 5 Tungsten Reference Compounds
Figure 5.34 – Theoretical L 3 Valence Band RIXS 2D contour plot for WCp*Me 4
5.9 Dimethoxyethane dichlorodioxotungsten
VI The W compound, WO 2Cl 2(dme), has W in a distorted octahedral geometry. The two oxide ions and the two oxygen atoms of dimethoxyethane ligand form the equatorial plane, with the 2 chlorine atoms at the apices 13 . The experimental HERFD and normal XANES are shown alongside the XANES spectrum calculated using FEFF in Figure 5.35. The XANES calculated by FEFF does not show the low energy shoulder on the main peak.
The density of states for WO 2Cl 2(dme) is shown in Figure 5.36. In the occupied density of states there is clearly a large amount of hybridisation with the W d orbitals, particularly with the p orbitals of the oxygen atoms in the dme ligand, but also with the oxide p orbitals and the C p orbitals. The unoccupied density of states is mostly W d, with a small amount of O p character.
113 Chapter 5 Tungsten Reference Compounds
Figure 5.35- L3 HERFD XANES at the valence band, normal fluorescence detected
XANES and XANES calculated by FEFF9 for WO 2Cl 2(dme)
Figure 5.36 - Density of states as calculated by FEFF9 for WO 2Cl 2(dme), O1 refers to oxide ligands, O2 refers to O atoms in dme ligand.
114 Chapter 5 Tungsten Reference Compounds
The experimental W L 3 VB RIXS plot for WO 2Cl 2(dme) is shown in Figure 5.37 and the FEFF9 simulation of this spectrum is shown in Figure 5.38. The energy transfer position of the experimental plot is well reproduced in the calculation, with the centre of the peak at 8.8 eV energy transfer. The density of states (Figure 5.36) shows a strong W d peak in the unoccupied region at ~2.5 eV, and with an energy transfer gap of ~8.8 eV expected we need to look at the occupied region we can see that the peak in the W dDOS at around -6.5 eV must be represent the molecular orbitals from which the emission occurs. There is particularly high hybridisation with the oxygen p orbitals from the d-orbitals. In addition we see a contribution from the W p orbitals as well as high contributions from the Cl and oxide O p orbitals.
Figure 5.37 - Experimental L 3 Valence Band RIXS 2D contour plot for WO 2Cl 2(dme)
115 Chapter 5 Tungsten Reference Compounds
Figure 5.38 - Theoretical L 3 Valence Band RIXS 2D contour plot for WO 2Cl 2(dme) as calculated by FEFF9
5.10 WCp*Cl 4
An L 3 VB RIXS spectrum was taken for WCp*Cl 4 but no calculations were performed due to lack of reliable structural data. The experimental spectrum is shown in Figure 5.39. The tungsten is in a +5 oxidation state.
The most intense peak is at 10209.1 eV incident energy (when calibrated to match the data from other beamtimes) and 8.4 eV energy transfer. There are two additional peaks at 10208.2 eV incident energy (after calibration) and 4.6 eV energy transfer, and 10207.3 eV incident energy (after calibration) and 6.2 eV energy transfer. The peaks at 6.2 eV and 8.4 eV appear to be on the main diagonal, if we take the position of the first peak on this diagonal (10207.3 eV IE and 6.2 eV ET) this is a similar energy transfer 5 position to the first peak of the d WCp*Me 4 complex, although the incident energy is lower by 1.1 eV. However it is not as clear from this spectrum where the position of the first peak should be taken from. Compared to the WCp*Me 4 RIXS spectrum (Figure
5.33) the spectrum for WCp*Cl 4 has features at lower energy transfers, indicating a smaller HOMO-LUMO gap.
116 Chapter 5 Tungsten Reference Compounds
Figure 5.39 - Experimental L 3 Valence Band RIXS 2D contour plot for WCp*Cl 4
5.11 Tungsten hexachloride
Tungsten hexachloride is a d 0 tungsten complex where the metal centre is in octahedral geometry. It damaged quickly in the beam, making RIXS impossible at ID26. 10 second XANES scans, of a short energy range, were continuously taken on the same position on the sample and the amplitude and features monitored for beam damage. The sample appeared to damage after 1-2 minutes in the beam. ID26 is a beamline with a very intense beam, of high flux. However HERFD data was acquired (scans lasting ~30 seconds) and is shown in Figure 5.40. The first peak at ~10202 eV is the elastic peak and the main peak follows at higher energies. There is significant splitting of the main peak, which corresponds to the splitting of the W d-orbitals (and any Cl p contributions at these energies).
117 Chapter 5 Tungsten Reference Compounds
Figure 5.40 - L3 VB HERFD spectrum of WCl 6
5.12 Discussion and comparison of Valence Band RIXS
The peak positions of the different W L 3 VB RIXS spectra shown in this chapter are summarised in Table 5.1 and some interesting comparisons can be made. The position of the centre of the first peak should give a guide to the oxidation state of the metal, but the type of ligand will also have an effect. We can look at the orbitals that contribute to the first peak to help investigate this, i.e. which metal orbitals are involved and if the primary contribution is metal or ligand orbitals. A general trend can be seen in the incident energy position of the first peak, with the exception of W(CO) 6, moving from 10207.7 eV for tungsten (IV) oxide, to 10208.4 eV for the W V complex, with the W VI compounds spanning a range from 10209.2 to 10210.0 eV. If we compare energy transfer positions for the first peaks we see again a diverse range for the W VI compounds, from 7.2 to 10.8 eV, which suggest ligand effects make a significant contribution.
118 Chapter 5 Tungsten Reference Compounds
Table 5.1 - Summary of experimental VB RIXS features for the tungsten reference compounds
Formal Centre of 2 nd Centre of Centre of 1st peak Compound and oxidation peak (Incident 3rd peak (Incident energy, geometry state of energy, energy (IE, ET) / energy transfer) / eV tungsten transfer) / eV eV
Li 2WO 4 +6 10209.6, 8.7 10212.5, 10.6 Tetrahedral
WO 3 Distorted +6 10210.0, 8.8 10214.0, 11.6 octahedral
WO 2 10210.1, Distorted +4 10207.7, 2.9 10211.1, 5.8 10.1 octahedral
WOCl 4 +6 10210.0, 7.2 - Square pyramidal
WO 2Cl 2(dme) Distorted +6 10209.8, 8.8 10212.0, 10.0 octahedral
t t W(NH Bu) 2(N Bu) 2 Distorted +6 10209.2, 7.2 10212.2, 9.5 tetrahedral
WCp*Me 4 +5 10208.4, 5.8 10212.4, 9.2 Square pyramidal
W(CO) 6 0 10209.6, 9.0 10211.9, 10.2 Octahedral
We see a very similar position for the first peak for lithium tungstate and tungsten (VI) oxide, both W VI with direct bonding to O, but with tetrahedral and octahedral geometry respectively. The energy transfer position of the first peaks for these two compounds differs by only 0.1 eV, and the incident energy position by 0.4 eV. However we can see that the splitting between the first and second peaks differs significantly between them.
119 Chapter 5 Tungsten Reference Compounds
For lithium tungstate the splitting is 2.9 eV in the incident energy direction and 1.9 eV in the energy transfer direction. For tungsten trioxide the splitting is 4 eV incident energy and 2.8 eV energy transfer. We would expect to see greater splitting for the octahedral complex than the tetrahedral, particularly along the diagonal (corresponds to HERFD XANES) as the absorption process involves the transition of a 2p electron into the unfilled 5d states. The crystal field splitting for an octahedral complex will be larger than that of a tetrahedral complex, although the WO3 octahedra are distorted so the ratio will differ from that of a perfect geometry.
IV If we then compare the energy transfer positions of WO 3 and Li 2WO 4 with WO 2, a W compound, we see that the first two peaks for WO 2 are at a lower energy than the first peaks for the two W VI compounds. So, when considering the same ligands, the energy transfer is clearly higher for W atoms with a higher charge. This is logical as the bonding orbitals will be pulled closer to a metal centre with a higher charge, therefore increasing the energy transfer gap. The energy transfer gap appears to further decrease as the ligands become less electronegative, for example moving from WO 3 VI VI and Li 2WO 4 (both W ) to WOCl 4 (also W ) the energy transfer position moves to 7.2 eV from 8.7 to 8.8 eV.
Tungsten hexacarbonyl does not appear to follow the trends shown by the other compounds, with the position of the first peak being 10209.6 eV incident energy and 9 eV energy transfer, consistent with WVI compounds with particularly electronegative ligands. The CO ligands will have pi back-bonding as well as sigma donor effects which will affect the position of the edge. It has been seen in previous studies 14 that the origin of the electron density in the bond, i.e. metal or ligand, has a significant effect on the edge position.
Using the trends shown for these reference compounds it should be possible to apply them to RIXS spectra of more unknown compounds or to look at RIXS spectra during reactions.
5.13 L beta 2 RIXS for tungsten reference compounds
A series of L beta 2 RIXS were acquired for the above reference compounds at the SuperXAS beamline at the SLS. However they showed little difference between the different compounds. Lbeta 2 RIXS were not the original aim of the beamtime but we were restricted by equipment problems at the beamline. For L beta RIXS we take
120 Chapter 5 Tungsten Reference Compounds
scans around the L β2 emission line whilst moving stepwise across the L3 absorption edge. So for the x-ray absorption we look at a 2p to 5d transition, and for the emission a 4d to 2p transition. Below are a selection of examples; WCl 6 (Figure 5.41), WCp*Me 4
(Figure 5.1) and W(CO) 6 (Figure 5.42), so covering a range of oxidation states and ligand types, but showing little difference in the RIXS planes. We were able to take a
RIXS scan of WCl 6 at the SLS, but not the ESRF, which is probably due to the much less intense beam. We would expect to see differences between the samples, as they cover a range of geometries, ligand types and oxidation states. The differences in the
L3 absorption edge have been demonstrated previously in this chapter, for the W L 3 VB RIXS and HERFD.
However, when looking at the core hole lifetime broadenings for the intermediate and final core holes in the RIXS this shows that the intermediate (L 3) core hole has a lifetime broadening of 4.98 eV, and the final core hole (N 4, N 5) has a lifetime broadening of around 4 eV 15 . Therefore this is unlikely to be sufficient to resolve fine structure in the RIXS plane. RIXS resolution is governed by the final state corehole lifetime, which here will be greater than for the VB RIXS, as the final state has 4d lifetime broadening whereas the VB RIXS do not have a corehole in the final state. The total experimental resolution for this experiment was 1.9 eV. Therefore it appears that the similarity between the spectra is due to the lifetime broadening which has resulted in spectra without sufficient resolution to distinguish the features.
Figure 5.41 - Lbeta2 RIXS spectra for WCl 6 (left) and WCp*Me 4 (right) with energy transfer plotted against incident energy. The areas in red designate high intensity at the detector
121 Chapter 5 Tungsten Reference Compounds
FEFF9 simulations were calculated for a selection of these compounds giving fair agreement in the energy transfer position of the peak, one example for tungsten hexacarbonyl is shown above in Figure 5.42. Removing experimental broadening in the calculation did not have an effect on the shape of the RIXS, confirming that the lifetime broadening is the main factor in the lack of energy resolution.
Figure 5.42 - Lbeta2 RIXS spectrum for W(CO) 6 with energy transfer plotted against incident energy. Experimental plot is on the left, FEFF calculation on the right. The areas in red designate high intensity at the detector.
5.14 Conclusions In conclusion this series of valence-to-core RIXS shows a clear sensitivity to oxidation state and ligand type. An extension to FEFF9 can be used to simulate the RIXS data but as this is still in development it is not currently reliable enough to calculate spectra for unknowns. Too many parameters need to be changed to fit each individual reference compound. However once a reasonable agreement is found between the experimental and theoretical spectra the density of states produced in the same calculation can be used to assign features in the RIXS. The series of reference materials with different ligand types, oxidation states and geometries show clear trends for incident and emissions energies which enables us to gain insights into these factors from the RIXS spectra. From the RIXS spectra we can also make inferences to the nature of the materials, i.e. metallic, conducting, semi-conducting, insulating.
122 Chapter 5 Tungsten Reference Compounds
5.15 References
1 Kas, J. J.; Rehr, J. J.; Soininen, J. A.; Glatzel, P., Phys. Rev. B ., 2011 , 83, 235114.
2 Palmer, D. J.; Dickens, P. G., Acta Cryst. B , 1979 , 35 , 2199.
3 Gulino, A.; Parker, S.; Jones, F. H.; Egdell, R . G.; Chem. Soc., Faraday Trans ., 1996 , 92 , 2137
4 Bolzan, A.; Kennedy, B.; Howard, C., Aus. J. Chem. , 1995 , 48 , 1473.
5 Yamazoe, S.; Hitomi, Y.; Shishido, T.; Tanaka, T., J. Phys. Chem. C ., 2008 , 112, 6869.
6 Akasura, H.; Shishido, T.; Yamazoe, S.; Teramura, K.; Tanaka, T., J. Phys. Chem. C., 2011 , 115 , 23653.
7 Loopstra, B. O.; Boldrini, P., Acta. Cryst., 1966 , 21 , 158.
8 Smolentsev, N.; Sikora, M.; Soldatov, A. V.; Kvashnina, K. O.; Glatzel, P., Phys. Rev. B, 2011 , 84 , 235113.
9 Hess, V. H.; Hartung, H., Z. Anorg. Allg. Chem., 1966 , 344 , 157
10 Choujaa, H.; Cosham, S. D.; Johnson, A. L.; Kafka, G. R.; Mahon, M. F.; Masters, S. L.; Molloy, K. C.; Rankin, D. W. H.; Robertson, H. E.; Wann, D. A., Inorg. Chem. , 2009 , 48, 2289.
11 Kohler, K., Steiner, A., Roesky, H. W., Z. Naturforsch. B. Chem. Sci., 1995 , 50 , 1207
12 Liu, A. H.; Murray, R. C.; Dewan, J. C.; Santarsiero, B. D.; Schrock, R. R. J. Am. Chem. Soc., 1987 , 109 , 4282
13 Dreisch, K.; Andersson, C.; Stalhandske, C., Polyhedron, 1991 , 10, 2417.
14 Tromp, M.; Moulin, J.; Reid, G.; Evans, J., Hedman, B.; Painetta, P., Eds. Amer Inst Physics: Melville, 2007 , 699.
15 Nyholm, R.; Martensson, N., Phys. Rev. B , 1987 , 36 , 20.
123
124 Chapter 6 Rhenium reference compounds
6 Rhenium Reference Compounds 6.1 Introduction In this chapter the high energy resolution fluorescence detected (HERFD) X-ray Absorption Near Edge Structure (XANES) and valence band resonant inelastic x-ray scattering (VB RIXS) spectra for a series of rhenium reference complexes will be discussed. The data was collected at ID26 at the European Synchrotron Radiation Facility (ESRF) in France for the Valence Band studies.
As described in Chapter 5 comparatively few XANES and RIXS studies exist for 5d metals compared to their 3d counterparts. In this chapter we aim to look in more detail at the electronic and geometric effects which govern the shape of these L 3 rhenium XANES spectra. By using the HERFD technique we can uncover features that were previously unresolved in normal XANES. In addition we have the Valence Band RIXS for the same compounds; this enables us to probe the region around the valence band in more detail.
The HERFD spectra were collected by recording a XANES spectrum at a specific emission energy, in this case 10533 eV, in order to reduce the life time broadening of the core hole. A secondary monochromator set at this energy was used to select the emission energy at which the absorption spectra would be collected. In addition to this there was a fluorescence detector set up behind the analyser crystals (i.e. the secondary monochromator) which gives us the more usual, fluorescence detected, XANES. The overall energy resolution of the RIXS planes was 1.1 eV. The intermediate core hole broadening is 5.04 eV but the final excited state does not have a core hole, hence a long life time 4.
Theoretical spectra were calculated using the real space Green’s function (RSGF) code FEFF9 1, which uses a new theoretical treatment of RIXS based on a real-space multiple-scattering Green’s function formalism and a quasiboson model Hamiltonian to account for the single-particle spectrum and multi-electron excitations respectively. This new method uses a convolution of the XAS and XES calculations for the complexes based on their crystal structures to calculate the RIXS planes.
6.2 Rhenium (IV) oxide Rhenium (IV) oxide is orthorhombic with a Pbcn space group 2 and the rhenium atoms are in octahedral geometry, as shown in Figure 6.1. The structure is a disorted rutile
125 Chapter 6 Rhenium reference compounds where each rhenium is co-ordinated to 6 oxygen atoms and each oxygen atom is co- ordinated to three rhenium atoms 2. The rhenium atoms are in a +4 oxidation state, therefore it is in a 5d 3 electron arrangement.
Figure 6.1 - Structure of rhenium (IV) oxide viewed down the a axis, blue represents tungsten atoms and red represents oxygen atoms.
Shown below (Figure 6.2) is the HERFD spectrum obtained at ID26 for rhenium (IV) oxide which has as a comparison the normal XANES (i.e. at all emission energies) for rhenium (IV) oxide and the FEFF simulation. These were collected during the same experiment. It is clear that there is much greater energy resolution in the HERFD as evidenced by the greater number of features visible and the sharpness of these features. The FEFF calculation reproduces the splitting of the main peak which is not fully resolved in the HERFD. The first peak in the HERFD, at around 10534 eV is the elastic peak, which is not calculated in FEFF.
In Figure 6.3 the local density of states (LDOS) as calculated by FEFF9 are shown, the high level of Re d and O p hybridisation is obvious in the occupied states (below 0 eV), particularly the region below -5 eV. As the calculated XANES (in Figure 6.2) has broadening added in the splitting of the main peak here corresponds to the first three Re d features above the Fermi energy (0 eV) in the density of states plot.
126 Chapter 6 Rhenium reference compounds
Figure 6.2- HERFD XANES spectrum for ReO 2 taken at the valence band emission energy (10533 eV) alongside the normal XANES and the spectrum simulated using FEFF9
Figure 6.3 - Local density of states plot for ReO 2 as calculated by FEFF9
127 Chapter 6 Rhenium reference compounds
In Figure 6.4 we can see the experimental L 3 VB RIXS spectrum and the FEFF9 simulation of this spectrum is shown in Figure 6.5. The energy transfer positions of the peaks are well reproduced, but the relative incident energy positions are less so. The centre of the higher energy transfer peak lies at a higher incident energy than the lower energy transfer peak in the experimental plot, but in the FEFF calculation it lies at a lower incident energy. A similar result for the RIXS calculation is seen in the work of Smolentsev et al 4 where they use a full potential density functional theory (DFT) approach to calculate the L 3 VB RIXS spectrum for a selection of rhenium and tungsten oxides.
The main feature in the calculated RIXS plot is centred at ~5 eV energy transfer and is a broad peak in the diagonal direction (constant emission energy, CEE). It seems that the peak corresponds to electron promotion into the unoccupied Re d states between 1 and 4 eV in the density of states plot (Figure 6.3). The emission process, to fill the 2p core hole, will be from the highest occupied d-orbitals, and this is a band between around -1 and -4 eV. It is likely that the off-diagonal peak at ~12 eV could be due to an absorption transition from 2p to unoccupied 5d, but then emission from a slightly deeper filled orbital, in this case the peak between -6 and -8 eV in the DOS, which is of high oxygen p character.
Figure 6.4 – Experimental Re L 3 Valence Band RIXS 2D contour plot for ReO 2
128 Chapter 6 Rhenium reference compounds
Figure 6.5 - Theoretical L 3 Valence Band RIXS 2D contour plot for ReO 2 as calculated by an extension of FEFF9
6.3 Rhenium (VI) oxide Rhenium trioxide has a cubic unit cell, with space group Pm3m 3. It consists of corner sharing ReO 6 units, therefore the rhenium atom is in octahedral geometry. The structure is shown below in Figure 6.6.
Figure 6.6 - Structure of rhenium (VI) oxide3 viewed down the a axis, blue represents tungsten atoms and red represents oxygen atoms.
129 Chapter 6 Rhenium reference compounds
The experimental L 3 XANES and VB HERFD calculations are seen in Figure 6.7, along with the XANES spectrum calculated using FEFF9. No more features are seen in the HERFD than the XANES but the peak is sharper. The FEFF9 XANES calculation shows a high energy shoulder or splitting which is not visible in the experimental XANES.
Figure 6.7 - HERFD XANES spectrum for ReO 3 taken at the valence band emission energy (10533 eV) alongside the normal XANES and the spectrum simulated using FEFF9
The density of states (DOS) plot is shown in Figure 6.8 which we can correlate to the theoretical RIXS plot in Figure 6.10. If we compare the unoccupied density of states to the diagonal cut through of the RIXS plot we can see where the peaks originate from in terms of orbital contributions. It appears that the energy transfer position of the RIXS feature (in Figure 6.10) at around 4.5 eV corresponds to the gap between the first unoccupied Re d state at ~1.5 eV and the occupied peak at ~-2.5 eV with a large degree of O p character as well as Re p and d contributions. The unoccupied peaks at 1.5 and 3 eV in the DOS correspond to the main peak on the diagonal of the RIXS calculation, including the low energy shoulder. The peak at ~7 eV in the DOS therefore corresponds to the final, highest energy peak along the diagonal.
130 Chapter 6 Rhenium reference compounds
Figure 6.8 - Local density of states plot for ReO 3 as calculated by FEFF9
The experimental L 3 VB RIXS plot is shown in Figure 6.9 and can be compared to the FEFF VB RIXS calculation in Figure 6.10. The FEFF plot has more features than the experimental plot but the overall shape agrees. Also the FEFF calculation is similar in appearance to the DFT calculation of the ReO 3 VB RIXS plot in the work by Smolentsev et al (2011) 4. This suggests that the resolution of the experimental RIXS could be obscuring these features. The relative intensities of the peaks appears to be wrong, as in the experimental plot the RIXS feature is most intense at around 10 eV energy transfer, but the most intense feature is at 4.5 eV energy transfer. Overall the FEFF calculation seems to have the feature at an energy transfer position that is too low.
131 Chapter 6 Rhenium reference compounds
Figure 6.9 - Experimental L 3 Valence Band RIXS 2D contour plot for ReO 3
Figure 6.10 - Theoretical L 3 Valence Band RIXS 2D contour plot for ReO 3 as calculated by an extension of FEFF9
132 Chapter 6 Rhenium reference compounds
6.4 Mer -trichlorotri(methyldiphenylphosphine)rhenium In the Re III complex mer-trichlorotri(methyldiphenylphosphine) rhenium the metal is in distorted octahedral geometry and a 5d 4 configuration. The structure used in the 5 calculations was adapted from a close analogue in the literature . The L 3 VB HERFD for this complex is shown in Figure 6.11 and just shows one broad peak.
Figure 6.11 - HERFD XANES spectrum for ReCl 3(PPh 2Me) 3 taken at the valence band emission energy (10533 eV) alongside the normal XANES and the spectrum simulated using FEFF9
133 Chapter 6 Rhenium reference compounds
Figure 6.12 - Local density of states for mer -ReCl 3(PMe 2Ph) 3 as calculated by FEFF9
The density of states as calculated by FEFF9 is shown in Figure 6.12 and shows the different orbital contributions and hybridisation. If the diagonal features of the calculated RIXS plane in Figure 6.14 are compared to the unoccupied density of states we can see the origin of the RIXS features. The main peak, centred at ~4 eV energy transfer, in the RIXS plot is due to transitions into the unoccupied orbital at ~1.5 eV in the DOS, which is completely Re d in character. The high energy shoulder to this main peak, at ~7 eV energy transfer is therefore due to the peak at ~5 eV in the DOS, which is predominately Re d. The difference between the occupied and unoccupied states should relate to the energy transfer position. In this case the highest occupied state shows a high degree of hybridisation between Re d orbitals and P and Cl p orbitals.
In Figure 6.13 and Figure 6.14 we see the experimental L 3 VB RIXS plot and the FEFF simulation of the L 3 VB RIXS plot respectively, both for mer -ReCl 3(PMe 2Ph) 3. There is good agreement in the energy transfer position of the main peak. The low energy transfer peak visible in the theoretical plot may be present in the experimental plot, but obscured by the elastic peak. The centre of the main peak in the RIXS spectra lies at just below 5 eV energy transfer in both spectra. The overall shape of the calculated plot appears to give fair agreement with the experimental plot if we assume the features in the experimental plot are less resolved, although there is an extra small peak at ~2 eV energy transfer.
134 Chapter 6 Rhenium reference compounds
Figure 6.13 - Experimental L 3 Valence Band RIXS plot for ReCl 3(PMePh 2)3
Figure 6.14 - Theoretical L 3 Valence Band RIXS plot for
ReCl 3(PMePh 2)3 as calculated by FEFF9
135 Chapter 6 Rhenium reference compounds
6.5 Trichlorooxobis(triphenylphosphine)rhenium(V)
In trichlorooxobis(triphenylphosphine)rhenium the metal is in a +5 formal oxidation state, therefore d 2 configuration. The rhenium has distorted octahedral geometry 6. The
L3 VB HERFD is shown in Figure 6.15, showing one fairly broad peak after the elastic peak. Also shown is the normal fluorescence detected XANES and the XANES spectrum calculated by FEFF9. Due to the noise in the HERFD it is difficult to determine if the high energy shoulder present in the FEFF is in the experimental spectrum, but it is possible that it is there. If we consider the density of states (DOS) plot (Figure 6.16) with respect to the XANES, as both are from the same FEFF calculation, the XANES spectra shape will reflect the unoccupied density of states. From this we can see that the main Re d peak in the DOS (at ~1eV), into which the 2p electron will be excited, has a significant amount of hybridisation with oxygen p orbitals.
Figure 6.15 – L 3 HERFD XANES spectrum for ReCl 3O(PPh 3)2 taken at the valence band emission energy (10533 eV) alongside the normal XANES and the spectrum simulated using FEFF9
136 Chapter 6 Rhenium reference compounds
Figure 6.16 - Local density of states for ReCl 3O(PPh 3)2 as calculated by FEFF9
There is fair agreement between the experimental L 3 VB RIXS spectrum (Figure 6.17) and the FEFF calculation (Figure 6.18) but the energy transfer of the peak in the calculation is slightly too low. Also in the experimental spectrum there is some splitting to give an off-diagonal shoulder at lower energy transfer. In the FEFF calculation the peak does extend down to the correct energy transfer region but there is no splitting from the main peak. It is possible that there is an error in the Fermi level for the calculation as this was the main cause of difficulties in this case.
We can use the density of states (Figure 6.16) to look at the orbital contributions to the RIXS spectrum calculated by FEFF (Figure 6.18). When comparing the diagonal cut through the RIXS to the unoccupied states it seems that the Re d peaks from ~1 eV to ~6 eV all contribute to the RIXS spectrum. The energy transfer position of the peak is centred at ~5 eV and should correspond to the gap between the highest occupied and lowest unoccupied d states but in this case it is not so clear cut as the gap between the first peaks on either side of the Fermi energy (0 eV) is only ~1.5 eV. Therefore the gap seems to correspond to the broad occupied Re d state at around -4 eV.
The FEFF calculation for this structure was particularly difficult but when the code has improved this can be recalculated in order to gain better results.
137 Chapter 6 Rhenium reference compounds
Figure 6.17 - Experimental Re L 3 VB RIXS plot for ReCl 3O(PPh 3)2
Figure 6.18 - Theoretical Re L 3 VB RIXS plot for ReCl 3O(PPh 3)2 as calculated by an extension of FEFF9
138 Chapter 6 Rhenium reference compounds
6.6 Rhenium tris(o-phenylenediamide)
The geometry around the central rhenium atom in Re(o-(HN) 2C6H4)3 is almost perfect trigonal prismatic geometry 7 and the metal is in a +6 formal oxidation state, therefore it is a 5d 1 complex. The experimental HERFD and normal XANES spectra are shown in Figure 6.19 along with the FEFF calculation of the XANES. The FEFF spectrum shows good agreement with the shape of the first peak (after the elastic peak) in the HERFD.
Figure 6.19 - L 3 HERFD XANES spectrum for Re(o-pda) 3 taken at the valence band emission energy (10533 eV) alongside the normal XANES and the spectrum simulated using FEFF9
We can correlate the local density of states plot (Figure 6.21) to the calculated XANES spectrum. The unoccupied Re d states from around 0 to 5 eV in the DOS result in the shape of the XANES peak. The features are better resolved in the DOS plot because broadening has been applied in the XANES spectrum. It is unclear why there is not a gap in the dDOS, because this is a molecular compound and should therefore have discrete levels
139 Chapter 6 Rhenium reference compounds
Figure 6.20 - Local density of states for Re(o-pda)3 as calculated by FEFF9
The experimental (Figure 6.21) and theoretical (Figure 6.22) L 3 VB RIXS spectra show good agreement in both the energy transfer position and shape of the peak. The diagonal cut through of the RIXS plane corresponds to the same region of the DOS as mentioned above for the XANES calculation, i.e. the unoccupied region between 0 and 5 eV. The emission transition, to fill the 2p core hole from one of the highest occupied states, occurs from the filled d state at -4.5 eV. The gap between this and the lowest unoccupied states (0 to 5 eV) corresponds to the energy transfer gap between the broad RIXS feature and zero energy transfer.
140 Chapter 6 Rhenium reference compounds
Figure 6.21 - Experimental Re L 3 VB RIXS plot for Re(o-pda) 3
Figure 6.22 – Theoretical Re L 3 VB RIXS plot for Re(o-pda) 3 as calculated by an extension of FEFF9
141 Chapter 6 Rhenium reference compounds
6.7 Summary
In Table 6.1 the positions of the major peaks in the VB RIXS spectra for the rhenium reference compounds are summarised in order to identify any trends.
Table 6.1 - Summary of peak positions for experimental Re L 3 VB RIXS
Centre of 1st peak Centre of 2 nd peak Compound and Formal oxidation (Incident energy, (Incident energy, geometry state of rhenium energy transfer) / eV energy transfer) / eV
ReO 3 +6 10543.5, 10.8 Octahedral
ReCl 3O(PPh 3)2 10539.5, 7.0 (Main + 5 10540.4, 3.8 Distorted octahedral peak)
ReO 2 +4 10541.4, 5.8 10543.1, 10.1 Octahedral
mer -
ReCl 3(PMe 2Ph) 3 +3 10539.8, 4.7 Distorted octahedral
Re(o-pda) 3 +6 10543.1, 9.0 Trigonal prismatic
As with the tungsten spectra in Chapter 5 there is a clear trend with oxidation state in the incident energy position of the first peak, moving from 10539.8 eV for the Re III complex, to 10541.4 eV for the Re IV compound and 10543.3 eV on average for the Re VI compounds. This is what you would expect as a broad trend because as the charge on the rhenium nucleus increases more energy will be needed to excite a core electron. However this will depend on where the charge is in the final state, it could be more ligand based than metal, so the individual DOS need to be considered.
However ligand types also have an effect on the edge position, so not all fit the pattern. V For example the Re complex, ReCl 3O(PPh 3)2 has a slightly lower incident energy position than expected, based on oxidation state alone we would expect it to be ~2 eV higher. When looking at the density of states plot for ReCl 3O(PPh 3)2 compared to the other compounds the one major difference is the significant level of hybridisation with
142 Chapter 6 Rhenium reference compounds the O p orbitals in the first unoccupied peak. Because there seems to be a higher degree of O than metal orbital density in the bonding, the metal will experience less of the charge, therefore the effect on the incident energy is reduced 8,9.
The same trend seems to apply for the energy transfer positions of the first peak, with the lowest being 4.7 eV for the Re III complex, increasing to between 9.0 and 10.8 eV for the Re VI compounds. As greater differences are seen in the energy transfer positions it could be used to better determine charges than the incident energy position However a combination of the position in both axes is best, combining information from both the filled and unfilled density of states.
6.8 Conclusions
From the results obtained here it is clear that the L3 VB RIXS spectra are sensitive to the metal oxidation state, as well as to ligand type. From the density of states plots it can be seen that orbital hybridisation influences the charge felt by the metal, and therefore the edge position.
It seems that experimental broadening might be obscuring some of the features in the RIXS, which can be seen in the density of states and calculated RIXS plots, but appear as one broad peak in the experimental RIXS plot. The experimental broadening was ~1.1 eV for this experiment, whereas the calculations have all broadening removed.
These RIXS spectra have shown that detailed oxidation states or charges can be derived from the plots, and due to the high energy resolution and 2D information this is superior to normal XANES. This is of interest when looking at unknown materials, and it is possible to use RIXS to study systems in-situ due to the hard x-rays.
6.9 References
1 Kas, J. J.; Rehr, J. J.; Soininen, J. A.; Glatzel, P., Phys. Rev. B., 2011, 83 , 235114.
2 A. Magneli, A., Acta Chem. Scand., 1957, 11, 28.
3 Jorgensen, J. E.; Jorgensen, J. D.; Batlogg, B.; Remeika, J. P.; Axe, J. D., Phys. Rev. B, 1986, 33 , 4793.
143 Chapter 6 Rhenium reference compounds
4 Smolentsev, N.; Sikora, M.; Soldatov, A. V.; Kvashnina, K. O.; Glatzel, P., Phys. Rev. B, 2011 ,84 , 235113.
5 Mitsopoulou, C. A.; Mahieu, N.; Motevalli, M.; Randall, E. W., J.Chem. Soc., Dalton Trans., 1996, 4563.
6 Lebuis, A.-M.; Beauchamp, A. L., Can. J. Chemistry, 1993, 71 , 441.
7 Danopoulos, A. A.; Wong, A. C. C.; Wilkinson, G.; Hursthouse, M. B.; Hussain, B., J. Chem. Soc., Dalton Trans., 1990, 315.
8 Tromp, M.; van Bokhoven, J. A.; van Strijdonck, G. P. F.; van Leeuwen, P.; Koningsberger, D. C.; Ramaker, D. E., J. Am. Chem. Soc., 2005, 127 , 777.
9 Tromp, M.; Moulin, J.; Reid, G.; Evans, J., AIP Conf. Proc., Hedman, B.; Painetta, P., Eds. Amer Inst Physics: Melville, 2007 , 882 , 699.
144 Chapter 7 Tungsten catalysis
7 Tungsten Catalysis 7.1 Introduction In this chapter we discuss the work that was carried out on a catalytic system, with the aim of extending the techniques and computational methods previously used to study tungsten reference compounds to a real system in order to obtain new information. We are aiming to study the electronic properties of catalytic materials, i.e. single site homogeneous systems, in detail. Information on the orientation of molecular orbitals, the charge transfer between them, and their accessibility for reactants, is obtained, providing insights in the properties required for (enantio)selective catalysis.
Here we study homogeneous tungsten catalysts used in the selective dimerisation of α- olefins. The majority of the existing catalysts will produce branched molecules; however certain late transition metal complexes have been successful in linear dimerisation 1. There are far fewer examples for group 6 metals, with the chromium catalysed trimerisation of ethylene2 the most successful case previously. Recent developments have shown tungsten imido complexes to be highly selective and moderately active 3. More specifically tungsten mono(imido) complexes, when treated with Lewis acids of the form AlCl nRm, have been reported as efficient catalysts for the dimerisation of α-olefins 4. A similar example has also been reported for tungsten bis(imido) compounds 5. Whereas a large excess of activator is usual for oligomerisation catalysts here a specific amount of ~15 molar equivalents is needed, indicating a particular function in the catalysis in addition to simple activation.
Figure 7.1 - General reaction scheme for tungsten imido catalyst formed in-situ for the dimerisation of olefins
145 Chapter 7 Tungsten catalysis
The reaction is summarised in Figure 7.1. In these experiments 1-hexene was the substrate, triethylamine (TEA) was used as a base to mop up excess chloride ions, and ethyl aluminium dichloride was used as the Lewis acid, all based on optimised parameters from Hanton et al 6. Reactions were carried out with either one or two equivalents of aniline in order to look at both the mono and bis(imido) species.
There are two different mechanistic pathways possible for these dimerisation reactions, one being the migratory insertion of olefins into metal-hydrogen or metal-alkyl bonds followed by β-hydride elimination, typical for the late transition metal systems. The more likely method, given the high selectivity and the substitution pattern of the dimerised products, is a metallacycle mechanism. However no direct mechanistic studies (i.e. labelling experiments) exist as yet. This metallacycle hypothesis has been examined in a theoretical study 7 using density functional theory (DFT) which presents the likely mechanism, however with no experimental data this cannot be proven.
There were a number of experimental difficulties with the systems chosen to study which has limited the data which we were able to gather. The tungsten species has limited solubility, as do the side products from the formation of the active species, so when at relatively high concentrations, as needed for XAFS studies, there was a tendency for precipitate to be formed after the addition of the final reagent. This was less of a problem for the RIXS studies as we were using a powerful x-ray beam and any glitches in the data were less problematic overall. Due to apparent beam damage the scans were limited to 1 hour maximum for obtaining a full RIXS plane. This was a compromise between the higher quality data of long acquisition times and the degradation of the samples. We originally hoped to obtain RIXS planes of the reaction intermediates by using a continuous flow cell to monitor a fixed point in the reaction. However due to the formation of precipitates we were unable to do this due to blockages.
The EXAFS data was not useful due to the glitches in the data resulting in poor quality spectra coupled with the fact that the tungsten was surrounded by light scattering elements. The solutions were all dark brown in colour ruling out UV Visible spectroscopy.
The reported high energy resolution fluorescence detected (HERFD) X-ray Absorption Near Edge Structure (XANES) and valence band resonant inelastic x-ray scattering
146 Chapter 7 Tungsten catalysis
(VB RIXS) spectra were collected at ID26 at the European Synchrotron Radiation Facility (ESRF) in France.
The HERFD spectra were collected by recording a XANES spectrum at a specific emission energy, in this case the L α1 emission line (8398 eV), in order to reduce the life time broadening of the core hole. A secondary monochromator set at this energy was used to select the emission energy at which the absorption spectra would be collected. In addition to this there was a fluorescence detector set up behind the analyser crystals (i.e. the secondary monochromator) which gives us the more usual, fluorescence detected, XANES. The experimental resolution was between 1.5 and 1.7 eV for the VB RIXS.
7.2 Experimental Results – VB RIXS
7.2.1 Experimental L 3 VB RIXS spectra
In Figure 7.2 we see on the left the L 3 valence band resonant inelastic scattering (VB RIXS) spectra for the reaction mixture before addition of the ethyl aluminium dichloride activator and on the right the end point of the same reaction (from a fresh starting solution to avoid the damage caused by the previous scan, and after stirring for ~1 hour). We initially hoped to take a RIXS spectrum of the species upon addition of the aluminium activator, using a continuous flow cell to monitor the same point in the reaction. However due to the precipitate formed during the reaction this was not possible. These RIXS plane took less than an hour each to obtain.
We can see from Figure 7.2 that the main difference between the starting species and the end species in the decrease in energy transfer, the peak being centred at around 9.65 eV energy transfer for the starting species but for the end species the peak is at around 5 eV energy transfer. The same trend is apparent in t Figure 7.3 (2 equivalents aniline used per WCl 6), Figure 7.4 (1 equivalent 4- butylaniline t used per WCl 6) and Figure 7.5 (2 equivalents 4- butylaniline used per WCl 6).
The only major difference we see in these spectra is the splitting of the RIXS feature in t the starting solution of WCl 6 and 2 equivalents of 4- butylaniline in Figure 7.5, but it is not clear what causes this. All measurements should have the same experimental resolution and be approximately the same concentration in solution, so it is unlikely that this double feature is simply unresolved in the other spectra. Although, from the reference studies in Chapter 5, we see similar energy transfer positions for systems with chlorine ligands and systems with imido ligands we would not expect to see no
147 Chapter 7 Tungsten catalysis difference between the mono(imido) and bis(imido) species. It is possible that with an overall resolution of around 1.7 eV for these experiments any differences are not resolved. Although unexpected it is also possible that the same species are formed, and therefore the stoichiometry of the reactants is not playing a significant role.
Figure 7.2 – Experimental L 3 VB RIXS spectra for Solution 1a: Starting solution of WCl 6 + 1 equivalent aniline + 3 equivalents of triethylamine in chlorobenzene and for Solution 1b: End species of Solution 1a + 1-hexene (1.6 mmol) + ethylaluminium dichloride solution (11 equivalents)
148 Chapter 7 Tungsten catalysis
Figure 7.3 - Experimental L 3 VB RIXS spectra for Solution 2a: Starting solution of WCl 6 + 2 equivalents aniline + triethylamine in chlorobenzene and for Solution 2b: End solution of Solution 2a + 1-hexene + 11 equivalents EADC in chlorobenzene
Figure 7.4 - Experimental L 3 VB RIXS spectra for Solution 3a: Starting solution of WCl 6 + 1 equivalent 4-tButylaniline + triethylamine in chlorobenzene and Solution 3b: End point of solution 3b + 11 equivalents EADC + 1-hexene.
149 Chapter 7 Tungsten catalysis
Figure 7.5 - Experimental L 3 VB RIXS spectra for Solution 4a: Starting solution of WCl 6 + 2 equivalents 4-tButylaniline + triethylamine in chlorobenzene and Solution 4b: End point of solution 4a + 11 equivalents EADC + 1-hexene in chlorobenzene
t Figure 7.6 - Experimental L 3 VB RIXS spectra for WCl 4(N-4 BuC 6H4) and Solution 5: t End species of WCl 4(N-4 BuC 6H4) precursor + 1-hexene (1.6 mmol) +EADC (11 equiv)
7.2.2 Summary table of experimental L 3 VB RIXS results The main features of the VB RIXS features for the catalysis experiments are summarised in Table 7.1, the RIXS contour plots themselves are shown in section
150 Chapter 7 Tungsten catalysis
7.2.1. These experiments were undertaken on two different beamtimes, along with other studies, so the incident energies of some experiments have been adjusted in this table due to the differences in absolute energy calibration. Therefore the values in this table may differ from the values shown in the contour plots.
From the table below we can see a clear trend. The pre-catalyst, formed from WCl 6 and aniline in chlorobenzene, has a peak at around 10210 eV incident energy and 8-9 eV energy transfer. The end species has a peak at around 10209 eV incident energy and around 5 eV energy transfer. Although the incident energy is very similar between start and end species, sometimes with a small decrease at the end point, there is a significant change in the energy transfer position. The peak moves from around 8 to 9 eV to around 5 eV over the course of the reaction.
When considering the W L 3 VB RIXS spectra from Chapter 5 the position of the starting solutions seems to be consistent with a W VI species in terms of incident energy. This is also true for the energy transfer position, which varies between 7.2 and 8.8 eV for the WVI compounds. From Chapter 5 it also appears that tungsten compounds with more electronegative ligands are at the higher end of this energy transfer range. This agrees with the proposed structure of the starting complex in solution, which is thought to be a tungsten (VI) mono(imido) tetrachloride species for systems with one equivalent of aniline 7,8.
Table 7.1- Summary of VB RIXS features for catalysis experiments
Incident Energy Figure Reaction Mixture Energy of Transfer of peak / eV peak / eV
Figure 7.2 Solution of WCl 6 (0.2 mmol) + 1 eq. aniline (Solution 10210.0 8.28 + 3 eq. TEA in 2 ml chlorobenzene 1a)
Figure 2 End species of WCl 6 (0.2 mmol) + 1 eq. (Solution aniline + 3 eq. TEA + 1-hexene (1.6 mmol) 10208.6 5.07 1b) + EADC (11 eq.) in 2 ml chlorobenzene
Figure 3 Solution of WCl 6 + 2 eq. aniline + 3 eq. (Solution 10210.0 8.20 TEA in 2 ml chlorobenzene 2a)
151 Chapter 7 Tungsten catalysis
Figure 3 End species of WCl 6 (0.2 mmol) + 2 eq. (Solution aniline + 3 eq. TEA + 1-hexene (1.6 mmol) 10209.3 5.16 2b) + EADC (11 eq.) in 2 ml chlorobenzene
Figure 4 Solution of WCl 6 (0.2 mmol) + 1 eq. 4- (Solution tButylaniline + 3 eq. TEA in 2 ml 10209.6 8.44 3a) chlorobenzene
End species of of WCl 6 (0.2 mmol) + 1 eq. Figure 4 4-tButylaniline + 3 eq. TEA + 1-hexene (Solution 10209.4 4.77 (1.6 mmol) + EADC (11 eq.) in 2 ml 3b) chlorobenzene
10209.2/ 8.07/ Figure 5 Solution of WCl 6 (0.2 mmol) + 2 eq. 4- 10210.2 8.98, (Solution tButylaniline + 3 eq. TEA in 2 ml CENTRE: CENTRE: 4a) chlorobenzene 10209.7 8.49
Figure 5 End species of WCl 6 (0.2 mmol) + 2 eq. 4- (Solution tbutylaniline + 3 eq. TEA + 1-hexene (1.6 10209.4 5.15 4b) mmol) + EADC (11 equivalents)
t Figure 6 WCl 4(N-4 BuC 6H4) solid 10209.2 8.94
t End species of reaction: WCl 4(N-4 BuC 6H4) Figure 6 precursor (0.2 mmol) + 1-hexene (1.6 10208.9 4.96 (Solution 5) mmol) +EADC (11 eq.) in 2 ml chlorobenzene
7.3 Experimental Results – HERFD reactions
In order to monitor changes during the reaction, L α1 HERFD spectra were taken every minute over the course of the reaction, and the results are shown in Figure 7.7. As they use a different emission line (Lα1) they can not be directly compared to the RIXS plots. These HERFD scans took less than 1 minute each to acquire, compared to ~1 hour for a full RIXS plane, so they are advantageous when looking at systems which change on a faster timescale (i.e. minutes not hours).
152 Chapter 7 Tungsten catalysis
This example is for the reaction of WCl 6 with 1 equivalent of aniline, 3 equivalents triethylamine and 11 equivalents ethylaluminium dichloride, to dimerise 1-hexene in chlorobenzene. The blue line is the starting reaction solution, before addition of the ethylaluminium dichloride (EADC). Unfortunately it was not possible to add the EADC on the beamline due to safety concerns so it was added in the chemical laboratory, which caused a minimum 5-10 minute delay between addition of the EADC and start of measurements. The other lines represent averaged spectra between 5 minutes and 1 hour after addition of EADC. It is clear that the only visible change occurs either with addition of the EADC or in the first 5 minutes after this, as all the HERFD spectra are identical when measurements recommenced. After addition of EADC the peak position moved to a lower energy (by 1.5 eV), and the high energy shoulder disappeared. Figure 7.8 shows a similar effect for the reaction with 1 equivalent 4-tbutylaniline, although the shoulder on the starting complex is less pronounced and the resolution is lower.
There is a clear difference in the shape of the peak and the position of the edge after addition of the EADC (edge shifts to lower energy and high energy shoulder disappears). As no further changes are seen after the ~5+ minutes that have elapsed between addition of the EADC and the next measurement it is unclear if the catalysis is already complete or if the dominant effect in the spectra is the effect of bonding with the EADC.
153 Chapter 7 Tungsten catalysis
Figure 7.7 – L 3 HERFD spectra of reaction of WCl 6 + 1 equivalent aniline + 3 equivalents triethylamine + 11 equivalents ethylaluminium dichloride to catalyse dimerisation of 1-hexene
t Figure 7.8 - Reaction of WCl 6 with 1 equivalent 4- Butylaniline and 3 equivalents TEA in chlorobenzene, to form pre-catalyst, along with 1-hexene (L 3 HERFD of starting solution, blue line). EADC was added and measurements were restarted after approximately 5 minutes (first HERFD shown in green, no further changes were visible after this).
154 Chapter 7 Tungsten catalysis
7.4 Other Experimental Results 7.4.1 Precursor vs. in situ The structure of the catalytic species generated in-situ has not been characterised, before or after addition of EADC, so as a comparison Figure 7.9 shows a solution of t WCl 4(N-4 BuC 6H4) in chlorobenzene alongside a solution of WCl 6 + 1 equivalent 4- tbutylaniline and 3 equivalents triethylamine in chlorobenzene. The two spectra are slightly noisy but appear to be identical, supporting the idea that the catalytic precursor formed in-situ is of the form WCl 4N=Ph.
t Figure 7.9 - L 3 HERFD spectra of WCl 4(N-4 BuC6H4) in chlorobenzene (blue) and of t WCl 6 + 1 equivalent 4- Butylaniline + 3 equivalents TEA + 1-hexene in chlorobenzene
7.4.2 Reaction with no substrate Figure 7.10 the VB RIXS spectrum for the end point of a reaction with no substrate is shown. The catalyst was formed in-situ using WCl 6, 1 equivalent aniline, 3 equivalents triethylamine and 11 equivalents ethylaluminium dichloride. The peak position is 10209.4 eV incident energy and 4.95 eV energy transfer, which is consistent with the peak position of the other end points, despite no dimerisation of 1-hexene occurring. Therefore it appears that the change is not related to the catalysis but instead to the addition of the EADC and its bonding with the tungsten molecules.
155 Chapter 7 Tungsten catalysis
Figure 7.10 - Experimental L 3 VB RIXS spectra of the end point (8 hrs) of a reaction without substrate (1-hexene). The catalyst was formed in-situ using WCl 6, 1 equivalent aniline, 3 equivalents triethylamine and 11 equivalents ethylaluminium dichloride.
7.4.3 Reaction with 6 equivalents aniline
Figure 7.11 - Experimental L 3 VB RIXS spectra for the starting solution of WCl 6 + 6 equivalents aniline (left) and the end point of the reaction initiated by addition of 1- hexene and 11 equivalents EADC (right)
156 Chapter 7 Tungsten catalysis
In Figure 7.11 the VB RIXS spectra for the start and end point of a reaction using 6 equivalents of aniline are shown. In the literature it is reported that this greatly restricts the catalysis from occurring 6. However the start and end point VB RIXS spectra show no clear difference from the catalysis spectra, indicating a similar end state species is formed. Since the catalysis mixtures are end states, all dimerisation might have happened and the same end state is left. So for future work this would be something to investigate.
7.5 Discussion and Conclusions From the results it is clear that the RIXS planes and HERFD spectra for all the end states show no significant differences between the different experiments. Based on the incident energy positions of the end state RIXS planes the oxidation state appears to be W VI , when compared to the W references in Chapter 5. However the energy transfer position is much lower so does not appear to agree with this, and lies somewhere between the ET positions for the W IV and W V compounds. However it has also been shown in Chapters 5 and 6 that ligand types can also influence the positions of the energy transfer, as it depends on the degree to which the charge is located on the metal or the ligand. Unfortunately we did not measure any reference compounds which contained both tungsten and aluminium for comparison. It is very likely that the ethyl aluminium dichloride is bonded to the W centre in some form.
Aluminium based activators play a role in many oligomerisation catalysts. A recent EXAFS study 9 probed the role of the aluminium activator for the selective olefin oligomerisation catalyst Mo(SNS)Cl 3, using trimethyl aluminium in place of MAO. It was found that in the end state the Cl ligands had all been replaced by methyl groups, which was also found in a study of analogous chromium systems 10 . It was also found in this study that Mo-Mo dimers, direct or chlorine bridged, did not appear to form. In a 27 45 11 similar EXAFS and NMR ( Al and Sc) study of Sc(SNS)Cl 3 complexes it appeared that after addition of AlMe 3 the chlorines were not substituted by the methyl groups, as with the group 6 analogues, but instead the Sc-Cl bond was activated by bridging to the
AlMe 3 group. This did not show an edge shift in the EXAFS after addition of the aluminium reagent, unlike the Cr and Mo which showed significant edge shifts. The spectra in this study also do not show a (significant) edge shift (i.e. in incident energy direction), which could suggest that the aluminium group is also bonded to the tungsten via a chlorine bridge. However this does not explain the shift in energy transfer. In the DFT studies by Tobisch 7 he proposes that one possible end state could be a chlorine
157 Chapter 7 Tungsten catalysis bridged W dimer; this could explain the big change in energy transfer position due to the close proximity of another W atom, whilst still having an incident energy position that suggests a +6 oxidation state. This would be an interesting route to investigate, perhaps by synthesising similar compounds for RIXS studies, or by theoretical studies when FEFF becomes more reliable for the study of 5d VB RIXS.
Unfortunately the FEFF simulations are not currently straight-forward enough for valence-to-core RIXS to reliably calculate spectra for intermediate species or end states. However this is in development and constantly improving, so in the near future it may be possible to re-examine these RIXS spectra using FEFF calculations based on models of potential structures. From this we could make clearer judgements about the nature of the Al bonding.
The synthesis and full structural characterisation of potential intermediates and end states would be one method by which to identify species involved in this catalysis. The same RIXS experiments could then be carried out on these well defined systems in order to compare to the unknowns that we have here, or to identify trends based on references with more similar ligand types or geometries to the catalytic species.
In summary there is not enough evidence from these RIXS studies to make any conclusions about the mechanism of this tungsten catalysis, although the results support the theory that the aluminium reagent plays a significant role. As this catalysis study was started towards the end of the PhD there was not time to explore a wider range of options. Future experiments and development of the theory should help us to gain more insight into the spectra presented here.
7.6 References
1 (a) Keim, W.; Hoffmann, B.; Lodewick, R.; Peuckert, M.; Schmitt, G.; Fleischhauer, J.; Meier, U., J. Mol. Catal., 1979, 6, 79. (b) Small, B. L.; Marcucci, A. J., Organometallics, 2001, 20 , 5738. (c) Small, B. L., Organometallics, 2003, 22 , 3178.
2 Carter, A., Cohen, S. A., Cooley, N. A., Murphy, A., Scutt, J., Wass, D. F., Chem. Comm., 2002 , 8, 858.
3 Olivier, H.; Laurent-Gérot, P., J. Mol. Catal. A- Chem, 1999, 148 , 43.
158 Chapter 7 Tungsten catalysis
4 Hanton, M. J., Tooze, R. P., WO 2005089940 (Sasol Technology (UK) Ltd), September 29, 2005.
5 Tobisch, S., Dalton Trans., 2008 , 16 , 2120.
6 Hanton, M. J.; Daubney, L.; Lebl, T.; Polas, S.; Smith, D. M.; Willemse, A., Dalton Trans., 39 , 7025.
7 Tobisch, S., Organometallics , 2007 , 26 , 6529.
8 Olivier, H.; Laurent-Gérot, P., J. Mol. Catal. A: Chem., 1999, 148 , 43.
9 Bartlett, S. A.; Wells, P. P.; Nachtegaal, M.; Dent, A. J.; Cibin, G.; Reid, G.; Evans, J.; Tromp, M., J. Catal., 2011, 284 , 247.
10 Moulin, J. O.; Evans, J.; McGuinness, D. S.; Reid, G.; Rucklidge, A. J.; Tooze, R. P.; Tromp, M., Dalton Trans., 2008, 1177.
11 Bartlett, S. A.; Cibin, G.; Dent, A. J.; Evans, J.; Hanton, M. J.; Reid, G.; Tooze, R. P.; Tromp, M., Dalton Trans., 2013, 42 , 2213.
159
160 Chapter 8 Conclusions
8 Conclusions and future work 8.1 Introduction The main aim of this project was to develop Resonant Inelastic X-ray Scattering (RIXS) spectroscopy, a relatively new and powerful x-ray technique, as a tool in homogeneous catalysis, to provide valuable insights into the (changing) electronic and geometric properties of the catalytic centre. Core-to-core RIXS experiments were carried out for a series of molybdenum reference materials, and valence-to-core RIXS experiments for a series of tungsten and rhenium reference compounds, with a range of oxidation states, geometries, and ligand types. Theoretical calculations were performed using FEFF9 in order to obtain an in-depth analysis of the electronic structure around the metal. These techniques were then applied to an active homogeneous tungsten catalyst.
8.2 Benefits of RIXS In Chapter 4 we have seen that from the experimental molybdenum 2p3d RIXS plots, we can obtain greater spectral resolution compared to the L 3 fluorescence yield XANES and as a consequence can provide more accurate oxidation state determination from the position of the first peak in the 2D plane. It also reveals additional off diagonal features that are not visible with XANES, which give insights into the geometry of the complex. The splitting of the main peak along the diagonal provides direct d orbital information, which can be related to the crystal field splitting.
The series of L 3 valence band (VB) RIXS for the tungsten and rhenium reference materials show clear trends for incident and emissions energies as a function of ligand type, oxidation state and geometry. This gives us insights into these factors from the RIXS spectra. From the RIXS spectra we can also make inferences to the nature of the materials, i.e. if they are metallic, conducting, semi-conducting or insulating. The derivation of oxidation states and charge transfers from the spectra appears to be superior to normal XANES, due to both the higher experimental resolution and the fact that the trends in oxidation state exist for both the incident energy and energy transfer directions (with the energy differences for the latter being larger).
We can also compare the high energy resolution fluorescence detected (HERFD) XANES to normal XANES, and it is clear that by reducing the lifetime broadening by using a specific emission line we see much clearer spectral resolution. In the spectra in Chapters 5, 6 and 7 the HERFD appear sharper than the normal XANES, and in some cases show features that are not visible in normal XANES plots.
161 Chapter 8 Conclusions
The RIXS techniques established in this project can now be applied to a wide range of materials and catalysts in order to obtain detailed electronic and geometric information under in situ and operando conditions. For unknown structures and changing materials the RIXS planes can be used as a direct measure for accurate oxidation states, geometry, ligand types and crystal field splitting.
One experimental consideration must always be the increased time and complexity of RIXS measurements in comparison to HERFD, when the RIXS plots often do not show anymore information. However as seen in sections 4 and 5 we do sometimes see off- diagonal features which may be of interest when considering the electronic structure of materials. So a good compromise may be to carry out HERFD scans of all materials and points of a reaction, then first assess these to see when a RIXS scan may be useful.
8.3 Comparison of experiment and theory FEFF9, an ab initio multiple scattering approach to calculating various x-ray spectra for clusters of atoms, was used to calculate the x-ray absorption near edge structure (XANES) and RIXS spectra in this work. FEFF requires a full 3D structure to run calculations, so a known or proposed structure is the main input. RIXS spectra can be calculated using a new extension of FEFF9, which is still in development, therefore there are still some problems with the code.
The density of states (DOS) are calculated alongside any absorption or emission spectra, and can be used to assign orbital contributions to spectral features. The DOS was calculated for all the reference compounds, with known structures and geometries, and shows what electronic information can be obtained. In the future we should be able to extend this to catalytic intermediates or end states. By looking at the trends in the RIXS planes for a number of well defined references we can derive possible ligand types, oxidation states and geometries. Based on this we can make models which can be used to calculate DOS and RIXS spectra in order to compare to the experimental RIXS spectra for validity. We could then use the DOS to derive detailed electronic information about the intermediates.
The core-to-core molybdenum 2p3d spectra can be simulated well using FEFF9, with good agreement in number and relative positions of peaks, since multiplets plays little
162 Chapter 8 Conclusions to no role in these spectra, with their current resolution. However the intensity ratio is not always well reproduced and the exact reasons are so far unclear.
For the valence-to-core (i.e. valence band) tungsten and rhenium RIXS in chapters 5 and 6 the FEFF simulations were often less successful. For example tungsten (VI) oxide gives poor agreement in the splitting of the peaks, but for tungsten (IV) oxide the splitting is in good agreement between the theory and experiment. Tungsten hexacarbonyl also gave poor results, which is not completely unexpected as FEFF does not always reproduce charge redistributions and π systems well. For some materials it is thought that there are full potential issues. FEFF uses a muffin-tin approximation in its treatment of atomic potentials, which generally gives good results but not in all cases. In these cases it is possible that use of a density functional theory approach, which uses a more sophisticated treatment of the potentials, might be more successful. The potential of ground state DFT calculations to simulate XES and VB RIXS spectra has been shown in the literature to be successful in some cases 1.
8.4 Catalysis
For the catalysis L 3 VB RIXS spectra were obtained for the start and end points, as well as HERFD taken during the reaction. The RIXS spectra showed a shift in energy transfer position from 8-9 eV for the starting species to ~5 eV for the end species. There were no significant differences between the different start or end points, i.e. upon change of reaction stoichiometry or additional substituent onto the aniline. There was a delay of ~5 minutes in between adding the aluminium activator to the starting solution, and taking the first HERFD spectrum in these in-situ experiments, and any change happened within these first 5 minutes.
When comparing to the W L 3 VB RIXS spectra for the references (Chapter 5) with the spectra for the catalysts (Chapter 7) the position of the starting solutions seems to be consistent with a W VI species in terms of incident energy and the energy transfer. This is also true for the energy transfer position, which varies between 7.2 and 8.8 eV for the W VI reference compounds. From Chapter 5 it also appears that tungsten compounds with more electronegative ligands are at the higher end of this energy transfer range. This agrees with the proposed structure of the starting complex in solution, which is thought to be a tungsten (VI) mono(imido) tetrachloride species for systems with one equivalent of aniline. HERFD spectra were taken of the catalyst
163 Chapter 8 Conclusions
t formed in-situ and the pre-made WCl 4(N-4 BuC6H4), and they seem to be identical, supporting this theory.
Based on the incident energy positions of the end state RIXS planes the oxidation state seems to be W VI , when compared to the W references compounds. However the energy transfer position is much lower so does not appear to agree with this, and lies somewhere between the ET positions for the W IV and W V compounds. However it has also been shown in Chapters 5 and 6 that ligand types can also influence the energy positions in the RIXS plane, depending on metal ligand hybridisation and the orbital character of the observed peak. The end states are the same for all the different reactions, even when catalysis should not take place, which indicates that the interaction of the starting complex with the ethylaluminium dichloride is the dominant effect in the RIXS. Unfortunately we did not measure any reference compounds which contained both tungsten and aluminium. It is very likely that the ethyl aluminium dichloride is bonded to the W centre in some form. The nature of this interaction of the Al reagent with the W centre is unclear. The large shift in energy transfer position should be due to a big change in the ligand type around the metal, causing a different charge distribution across the bond. One possibility is the formation of a Cl bridged W dimer. Based on examples in the literature 2 it does not appear that the Al reagent is alkylating the W, because this has been shown to cause a big edge shift, which is not seen in the incident energy position.
In summary there is not enough evidence from these RIXS studies to make any conclusions about the mechanism of this tungsten catalysis, although the results indicate that the aluminium reagent plays a significant role. Unfortunately the FEFF simulations are not currently reliable enough for valence-to-core RIXS to calculate reliable spectra for unknowns.
8.5 Recommendations for future work As it has been shown that accurate oxidation states and geometry information can be extracted from the Mo L α RIXS plots, it would be interesting to use this technique to study a catalytic system in situ /operando. However, due to the relatively low X-ray energies used, special experimental cells would have to be developed to allow these measurements in reduced atmosphere chambers. Although the experimental set-up would be challenging, Mo Lα RIXS is a powerful technique to probe the changing electronic properties of catalysts.
164 Chapter 8 Conclusions
The analysis of the core to valence RIXS was more challenging and the theoretical approaches are less well established. However they are improving constantly so in the near future it should be relatively simple to run FEFF calculations for VB RIXS to give reliable results. Once this is possible then we can look at extending these theoretical techniques to more ‘unknown’ samples, e.g. catalytic intermediates or end states where although the structure is not known a range of possibilities can first be deduced. At the same time it would be useful to attempt to calculate these RIXS plots using different approaches, for example a full potential DFT code, in order to compare results. For some structures it seems that there were full potential issues with the FEFF calculation, i.e. the muffin tin approximation was not sufficient to give good results. In these cases we hope to commence testing of the use of full potential codes. However these DFT codes are also not well established for applications in VB RIXS, so this approach will also be challenging at first.
More work is needed on the catalytic system to gain any detailed insights and there are a number of potential experiments which could be undertaken, but they were not possible in the time frame of this work. In the DFT studies by Tobisch 3 he proposes that one possible species that may be formed is chlorine bridged W dimer, this would be an interesting route to investigate, perhaps by synthesising similar compounds for RIXS studies, or by theoretical studies when FEFF becomes more reliable for these spectra. In addition DFT full potential methods should be investigated for the simulation of these spectra, and this is something that will be probed in the near future once it is has been tested with the reference compounds.
It would be advantageous to synthesise a range of potential catalytic intermediates and end states, or at least W complexes that are as close to these structures as possible in terms of oxidation state, ligand type and geometry. With RIXS and HERFD reference spectra of these compounds it should be easier to make deductions about the RIXS spectra of our end states.
8.6 References
1 Smolentsev, N.; Sikora, M.; Soldatov, A. V.; Kvashnina, K. O.; Glatzel, P., Phys. Rev. B, 2011, 84 , 235113.
165 Chapter 8 Conclusions
2 For example: Bartlett, S. A.; Wells, P. P.; Nachtegaal, M.; Dent, A. J.; Cibin, G.; Reid, G.; Evans, J.; Tromp, M., J. Catal., 2011, 284 , 247.
3 Tobisch, S., Organometallics 2007, 26 , 6529.
166